Economic Variables and their Relationship to the Returns ...
113
University of Cape Town Economic variables and their relationship to the returns of listed and unlisted commercial properties in South Africa by Che Wei Joey Lin LNXCHE003 A minor dissertation presented to the Department of Construction Economics and Management in partial fulfilment of the requirements for the degree MSc in Property Studies August 2014
Transcript of Economic Variables and their Relationship to the Returns ...
Univers
ity of
Cap
e Tow
n
Economic variables and their relationship to the returns of listed and unlisted commercial properties in South Africa
by
Che Wei Joey Lin
LNXCHE003
A minor dissertation presented to the Department of Construction Economics and
Management in partial fulfilment of the requirements for the degree MSc in Property Studies
August 2014
The copyright of this thesis vests in the author. No quotation from it or information derived from it is to be published without full acknowledgement of the source. The thesis is to be used for private study or non-commercial research purposes only.
Published by the University of Cape Town (UCT) in terms of the non-exclusive license granted to UCT by the author.
Univers
ity of
Cap
e Tow
n
ii
Plagiarism Declaration
1. I know that plagiarism is wrong. Plagiarism is using another’s work and pretending that
it is one’s own.
2. I have used the Harvard convention for citation and referencing. Each contribution to,
and quotation in, this dissertation from the works of other people has been attributed,
and has been cited and referenced.
3. This dissertation is my own work.
4. I have not allowed, and will not allow, anyone to copy my work with the intention of
passing it off as his or her own work.
5. I acknowledge that copying someone else’s assignment or essay or part of it is wrong,
and I declare that this is my own work.
____________________________
Che Wei Joey Lin
iii
Abstract
The purpose of this research is to investigate the relationship between unlisted and listed
commercial property returns and the macroeconomic factors identified, which are the stock
market, economic activity, inflation and interest rates, in South Africa for the period from 1995
to 2013 (for unlisted properties) and from 2002 to 2013 (for listed properties). It is commonly
understood that relevant macroeconomic variables impact asset prices; it is therefore easy to
see why it is important to examine the dynamic interactions between the macroeconomic
variables and property returns.
Previous studies identified stock market performance, economic growth, interest rate and
inflation as significant macroeconomic variables. The empirical research in this work is
conducted using regression and vector autoregression (VAR) methodologies consistent with
prior studies. Regression analysis considers the statistical dependence of the dependent
variable on one or more explanatory variables. VAR analysis permits inferences to be drawn
about how a particular variable helps to explain property returns and to see how a shock from
the same variable affects that return.
The work concluded that unlisted property has insufficient historical data to perform the
relevant statistical testing. It also established that unlisted property has shown a high
correlation (69%) to listed property. Finally, for listed property it was determined that interest
rates were found to be a significant negative variable. This result was consistent with the
impulse response analysis conducted. Variance decomposition also showed that the interest
rate variable explained almost 49% of the volatility of listed property. No other economic
variables identified in this work were found to be statistically significant.
This research is the first of its kind relating to commercial property in South Africa. The findings
of this research reaffirm the theoretical argument that the relationship between interest rates
and returns of commercial property is negative. The findings of this research are of significance
to investors, analysts and policymakers wishing to acquire a better understanding of this
market.
iv
Dedication
To my father, mother, wife and son for their support in my life and towards my dissertation.
Acknowledgements
Dr Ryan Kruger, Department of Finance and Tax, University of Cape Town
Professor Francois Viruly, Department of Construction Economics and Management, University
of Cape Town
Kathleen Evans, Department of Construction Economics and Management, University of Cape
Town
Kevin Kotze, School of Economics, University of Cape Town
Lara Kruger, School of Economics, University of Cape Town
Bureau of Economic Research, University of Stellenbosch
Stan Garrun, Investment Property Databank (IPD)
Niel Harmse, Investment Property Databank (IPD)
Zayd Sulaiman, Catalyst Fund Manager
v
Table of Contents
Plagiarism Declaration ................................................................................................................... ii
Abstract ........................................................................................................................................ iii
Dedication .................................................................................................................................... iv
Acknowledgements....................................................................................................................... iv
Table of Contents ........................................................................................................................... v
List of Figures ............................................................................................................................... vii
List of Tables ................................................................................................................................ vii
List of Equations ......................................................................................................................... viii
Abbreviations ............................................................................................................................... ix
1. Introduction ................................................................................................................... 1
1.1 Introduction and purpose of this chapter ............................................................................1
1.2 South African property market ...........................................................................................2
1.3 Listed property ..................................................................................................................2
1.3.1 Returns on listed properties ...............................................................................................4
1.3.2 Composition of listed properties ........................................................................................5
1.3.3 Regulation and structural change to REIT ............................................................................6
1.4 Unlisted property ...............................................................................................................7
1.5 Research problem statement .............................................................................................8
1.6 Research question..............................................................................................................8
1.7 Research aim .....................................................................................................................8
1.8 Research proposition .........................................................................................................9
1.9 Research objectives ...........................................................................................................9
1.10 Hypothesis .........................................................................................................................9
1.11 Methodology ................................................................................................................... 10
1.12 Justification of the research ............................................................................................. 10
1.13 Limitations of this research .............................................................................................. 10
1.14 Outline of the research report .......................................................................................... 11
2. Literature Review ......................................................................................................... 12
2.1 Stock market ................................................................................................................... 15
2.2 Economic growth and the property market ....................................................................... 17
2.3 Interest rates ................................................................................................................... 20
2.3.1 Types of interest rate proxies ........................................................................................... 23
2.3.2 Non-academic empirical studies ....................................................................................... 25
2.4 Inflation ........................................................................................................................... 28
2.4.1 Actual inflation ................................................................................................................ 28
2.4.2 Expected/unexpected inflation ........................................................................................ 30
2.5 Summary of the literature review ..................................................................................... 31
2.6 Other variables ................................................................................................................ 35
vi
2.7 South African literature .................................................................................................... 36
2.8 Summary and conclusion ................................................................................................. 40
3. Research Methodology ................................................................................................ 42
3.1 Statistical methodology ................................................................................................... 42
3.1.1 Regression analysis .......................................................................................................... 45
3.1.2 Vector autoregression (VAR) ............................................................................................ 47
3.2 Data and proxy selection .................................................................................................. 52
3.2.1 Unlisted properties .......................................................................................................... 52
3.2.2 Listed properties .............................................................................................................. 54
3.2.3 Comparison between the proxy of listed and unlisted properties ...................................... 56
3.2.4 Other proxies considerations............................................................................................ 58
3.3 Application of statistical methodology .............................................................................. 59
3.4 Summary ......................................................................................................................... 60
4. Results and Discussion ................................................................................................. 61
4.1 Unlisted properties (IPD) .................................................................................................. 61
4.1.1 Sample period analysis ..................................................................................................... 61
4.1.2 Descriptive statistics ........................................................................................................ 63
4.1.3 Regression analysis and vector autoregression ................................................................. 66
4.2 Listed (J256T, Property Loan Stocks) ................................................................................. 66
4.2.1 Sample period analysis ..................................................................................................... 66
4.2.2 Descriptive statistics ........................................................................................................ 68
4.2.3 Regression analysis .......................................................................................................... 71
4.2.4 Vector autoregression ...................................................................................................... 75
4.2.4.1 Variance decomposition ............................................................................................... 81
4.2.4.2 Impulse response ......................................................................................................... 82
4.3 Summary and discussion .................................................................................................. 86
5. Conclusion and Suggested Future Research .................................................................. 90
6. References and Bibliography ........................................................................................ 92
Appendix 1: Investment analysts’ views on listed property .................................................. 99
Appendix 2: Data used ....................................................................................................... 102
vii
List of Figures Figure 1: Market capitalisation of listed real estate chart book – October 2013 (SA REIT 2013c) 3
Figure 2: Investment rate vs borrowing rate (Bureau of Economic Research, University of
Stellenbosch, prepared by author) ............................................................................................... 25
Figure 3: Quarterly Return of PLS – August 2013 (SA Reit Association 2013a) ............................ 55
Figure 4: Returns of PLS, IPD and ALSI .......................................................................................... 57
Figure 5: Business cycle and ten-year bond rate over the sample period ................................... 62
Figure 6: Data of variables ............................................................................................................ 63
Figure 7: Business cycle and ten-year bond rate over the sample period ................................... 67
Figure 8: Data of Variables ............................................................................................................ 68
Figure 9: Testing the unit root of VAR .......................................................................................... 78
Figure 10: VAR residual ................................................................................................................. 79
Figure 11: Impulse response (Graph format) ............................................................................... 85
List of Tables Table 1: Return of listed property (INET BFA, calculated by author) ............................................. 4
Table 2: Market capitalisation of top REITs on the JSE (INET BFA) ................................................. 5
Table 3: Summary of literature review (showing inconsistency between period, frequency of
data, methodologies and results) ................................................................................................. 18
Table 4: Summary of relationships between economic growth and property returns ................ 19
Table 5: Summary of literature reviewed above (showing inconsistency between period,
frequency of data, methodologies and results ............................................................................. 22
Table 6: Summary of literature review (non-academic research) ................................................ 27
Table 7: Summary of literature reviews above (showing inconsistency between period,
frequency of data, methodologies and results ............................................................................. 29
Table 8: Summary of academic literature review (in alphabetical order) .................................... 31
Table 9: Summary of South African academic literature review .................................................. 39
Table 10: Summary of statistical methodologies to be applied ................................................... 43
Table 11: Summary of academic literature review (in alphabetical order) .................................. 43
Table 12: The variables and proxies used ..................................................................................... 54
Table 13: Summary of variables, proxies, period and source of data .......................................... 56
Table 14: Summary of variables, proxies, period and source of data .......................................... 56
Table 15: Descriptive statistics ..................................................................................................... 64
Table 16: Jarque–Bera test ........................................................................................................... 65
Table 17: Correlation matrix ......................................................................................................... 66
Table 18: Descriptive statistics ..................................................................................................... 69
viii
Table 19: Jarque–Bera test ........................................................................................................... 70
Table 20: Correlation matrix ......................................................................................................... 70
Table 21: Regression results after incorporation of autoregressive function (all variables) ....... 71
Table 22: Stepwise regression results........................................................................................... 72
Table 23: Regression results after incorporation of autoregressive function (two variables) ..... 73
Table 24: Assumptions for multiple regression ............................................................................ 74
Table 25: Dickey–Fuller test .......................................................................................................... 76
Table 26: VAR results .................................................................................................................... 77
Table 27: VAR residual portmanteau tests for autocorrelations .................................................. 80
Table 28: Variance decomposition ............................................................................................... 82
Table 29: Impulse response (Table format) .................................................................................. 83
Table 30: Summary of results ....................................................................................................... 86
Table 31: Summary of survey of analysts ..................................................................................... 99
List of Equations Equation 1: Capital Asset Pricing Model ....................................................................................... 16
Equation 2: Fisher theorem on deriving unexpected inflation ..................................................... 30
Equation 3: Regression ................................................................................................................. 46
Equation 4: Regression of Unlisted Property ................................................................................ 46
Equation 5: Regression of Listed Property.................................................................................... 47
Equation 6: General form of VAR.................................................................................................. 49
Equation 7: VAR of Unlisted Property ........................................................................................... 49
Equation 8: VAR of Listed Property .............................................................................................. 50
ix
Abbreviations
AIC Akaike’s information criterion
ALBI All-Bond Index
ALSI All-Share Index
APT arbitrage pricing theory
BER Bureau of Economic Research
CAGR compound annual growth rate
CAPM capital asset pricing model
FDW Fischer-DiPasquale-Wheaton (real estate model)
GDP gross domestic product
INET BFA a market data service provider
IPD Investment Property Databank
JSE Johannesburg Stock Exchange
NCREIF National Council of Real Estate Investment Fiduciaries
PLS property loan stock
PUT property unit trust
REIT real estate investment trust*
SA South Africa
UK United Kingdom
US United States of America
VAR vector autoregression
*This thesis refers only to Equity REIT.
1
1. Introduction
1.1 Introduction and purpose of this chapter
Property is an important real asset for investors because of its ability to diversify and provide an
inflation hedge that is better than most financial assets (Ilmanen 2012, p. 106). The phrase ‘to
be as safe as houses’ perpetuates the notion that property is among the least risky assets in the
mind of investors, but as Niall Ferguson (2008, p. 229) points out, ‘a bet on bricks and mortar is
very far from being as safe as houses’.
According to research commissioned by the Property Sector Charter Council, the South African
property sector is valued at approximately R4.9 trillion, with residential property estimated at
R3 trillion, commercial property at R780 billion, undeveloped land at R520 billion and
government property at R570 billion (Property Sector Charter Council 2012). As of 2011, the
real estate sector contributed 6.0% towards South Africa’s gross domestic product (GDP)
(Bureau of Economic Research 2011). Thus, it is an important sector in the South African
economy compared to agriculture at 2.2% and mining at 8.8%.
It is commonly believed that relevant economic variables impact asset prices (Chen, Roll & Ross
1986). Daily reports from financial media (newspapers, magazines, television and the internet)
tend to support the view that asset prices are influenced by a variety of events, some of which
seem to have more pervasive effects on asset prices than others.
There has been limited empirical research in South Africa on the linkages between property
returns and economic variables. Prior research focused on residential properties (see Standish,
Lowther, Morgan-Grenville & Quick 2005; Clark & Daniel 2006; Franken, Bloom & Erasmus
2011). This research will add to the body of knowledge and be of interest to investors, analysts
and policymakers.
The remainder of this chapter provides an overview of property markets in South Africa. It also
introduces the research problem statement and the research question; the aim, proposition,
2
and objectives of the study; and the proposed hypothesis, methodologies and justification for
the research. The chapter concludes with an outline of the research report.
1.2 South African property market
In South Africa, an investor can access property investments through two main methods,
namely:
1) via direct investment – whereby an investor purchases a physical property (such as an
office, house or apartment) or
2) through securitised instruments such as listed property stocks (Ilmanen 2012, p. 106).
1.3 Listed property
The South African listed property sector currently has two main structures, namely property
unit trusts (PUT) and property loan stocks (PLS). However, since 2013 the listed sector has
switched to real estate investment trust (REIT), and a number of funds are being converted to
this new structure. REIT provides tax certainty as it qualifies for the REIT tax dispensation as per
Section 25BB of the Taxation Legislation Amendment Bill (South African Government 2013) and
also provides investors with a similar structure to international standards.
The United States (US) is the leading market in securitised property. In 1961, the US introduced
the REIT structure. There is an abundance of historical data on the US market, which has
resulted in significant research in the US context. In South Africa, on the other hand, there has
been limited research in this sector (Payne 2003).
There are three types of REIT structures, namely equity, mortgage and hybrid (Payne 2003).
This research only discusses equity REITs, as none of the other products are available in South
Africa.
3
As of November 2013, there were 41 REITs (Business Day 2013) listed on the Johannesburg
Stock Exchange (JSE), with a market capitalisation of R236 billion (INET BFA1 2013).
The listed real estate sector has enjoyed tremendous growth over the last two decades. Market
capitalisation has grown from R13 billion in 2002 to R236 billion in 2013 (INET BFA; see Figure
1). At R236 billion, the listed property sector accounts for less than 3% of the total market
capitalisation of the JSE.
Figure 1: Market capitalisation of l isted real estate chart book – October 2013 (SA REIT 2013c)
1 INET BFA is a securities market data provider, created through the acquisition of Inet by McGregor BFA. http://www.inetbfa.com/
4
1.3.1 Returns on listed properties
Ilmanen (2012, pp. 102 and 111) noted that the long-run return of real estate in the US lies
between that of bonds and stocks, and the real ex ante real estate risk premium over the ten-
year treasury averaged 4% (from 1965 to 2006). Based on a review of prior research, Ilmanen
(2012, p. 111) concluded that the long-run real return of real estate reflects mainly the cash
yield, and real price growth is negligible; also it appears that the starting valuation matters for
research, as real estate is subject to the significant impact of boom-bust cycles and mean-
reverting valuations.
In South Africa, listed property outperformed both bonds and equities between August 2003
and August 2013 (see Table 1). The Property Loan Stock Index showed an annual return of
25.6% over a ten-year period. Over the same period, the JSE All-Share Index showed an annual
return of 19.8%. Bonds over this period yielded an average 9.2% annual return according to the
All-Bond Index.
Table 1 illustrates returns for the JSE’s property loan stock, the JSE’s All-Share Index (ALSI), and
the JSE’s All-Bond Index (ALBI):
Table 1: Return of l isted property (INET BFA, calculated by author)
Property Loan Stock
(Total Returns)
ALSI Index
(Total Returns)
ALBI
(Total Returns)
2003 – 2013 CAGR Return p.a. 25.6% 19.8% 9.2%
As illustrated in Table 1, the listed property sector has provided excellent returns over the last
ten years. Together with the steady increase in the number of listings and changes to the REIT
structure, South African and foreign investors are giving increasing attention to the real estate
sector (Smith 2013; Schnehage 2012).
5
1.3.2 Composition of listed properties
Commercial properties – consisting of office, retail and industrial properties – make up the
majority of the underlying property in the listed real estate sector in South Africa. According to
Stanlib (Anderson 2013a), South Africa’s listed property sector has only 1% exposure by value
to residential property, compared with about 11% in developed markets and 15% in emerging
markets. However, recent trends indicate that listed properties are interested in increasing
their investments in residential properties, for example, the recent acquisition by Arrowhead
Properties of Jika Properties’ residential portfolio for R406 million (Anderson 2013b). Thus,
listed property indices provide a good benchmark for listed commercial properties in South
Africa.
Currently, of the 41 REITs listed on the JSE, the top four dominate 46% of the market
capitalisation of the listed property sector (see Table 2).
Table 2: Market capitalisation of top REITs on the JSE (INET BFA)
Market Capitalisation as of Nov 2013 Billion ZAR % of Total
Growthpoint 46 20%
Redefine 28 12%
Hyprop 18 8%
Resilient 16 7%
Remainder 127 54%
SA REIT (J867) 236 100%
6
1.3.3 Regulation and structural change to REIT
Prior to the introduction of REIT, the listed real estate structure was comprised of property loan
stock (PLS) and property unit trusts (PUT). The difference is in the corporate structure and tax
structure.
A PUT is a collective investment scheme in property and is governed by the Collective
Investment Schemes Act and the Financial Services Board. PUTs have tax certainty, and the
income distributed to unitholders is not taxed in the PUT. It retains its nature and is taxed in the
hands of the unitholder according to their tax status.
A PLS is a property loan stock company, which has a share linked to a variable rate debenture.
PLSs have fewer restrictions than PUTs (for example, gearing is unlimited and they can invest in
other companies), but they do not have tax certainty (SA Corporate Real Estate Fund n.d.).
The planning process commenced in August 2006 with a REIT conference, followed by three
years of work between Property Loan Stock Association’s REIT committee and the South African
National Treasury. The result was the Taxation Legislation Amendment Bill released in February
2013. The bill contains the Section 25BB REIT tax dispensation, aligning the South Africa listed
real estate market with the internationally recognised global standard (SA REIT Association
2013b).
The requirements for a JSE-listed South African REIT are as follows:
own at least R300 million of property;
keep its debt below 60% of its gross asset value;
earn 75% of its income from rental or property owned or investment income from
indirect property ownership;
have a committee to monitor risk;
not enter into derivative instruments that are not in the ordinary course of business;
pay at least 75% of its taxable earnings available; and
make a distribution to its investors each year.
7
The change of a listing structure to REIT may change its current relationship in terms of
profitability. As pointed out by Graff (2001), investment companies can generate long-term per-
share earning growth in two ways: by investing in assets with growing earnings and by financing
investment portfolio expansion through reinvestment of retained earnings. Since real estate is
not a growth asset, REITs can grow per-share earnings only by reinvesting retained earnings to
expand underlying real estate portfolios. In the case of South Africa, as REITs are required to
pay out 75% of taxable ordinary income to shareholders, REITs are not growth stocks, but
cyclical income-generating assets with comparable investment characteristics to underlying
investment portfolios.
The conversion to the REIT structure has no impact on this research, as the data used for this
research were all pre-conversion structures.
1.4 Unlisted property
Although the market capitalisation of South African listed property is R236 billion (INET BFA),
the unlisted sector remains substantial. Investment Property Databank (IPD) estimates that
about 54% of the country’s professionally managed investment property is listed (Hedley 2013),
with the predominant players in the unlisted arena being either insurance companies or
pension funds (such as, Liberty Properties, Old Mutual Property, Sanlam Properties, the PIC,
Momentum Property Fund and Sasol).
Property economist Francois Viruly (as quoted in Hedley 2013) points out that some of the
prime properties in South Africa are majority-owned by unlisted players such as Sandton City
(by Liberties Properties), and Cavendish Square and Gateway (by Old Mutual). He believes that
once the new REIT structure has ‘bedded itself down’, some of the unlisted funds could list in
the next few years.
8
1.5 Research problem statement
The problem to be examined in this study is stated as:
International studies have suggested that there are relationships between macroeconomic
variables and the performance of the commercial property sector (see Chen, Peng, Shyu & Zeng
2012; Downs, Fung, Patterson & Yua 2003; Payne 2003). This research considers whether the
conclusions of these studies have relevance in the South African context. To date, no similar
research has been comprehensively conducted in South Africa.
1.6 Research question
The research questions to be addressed are as follows:
a) Is there a significant relationship between macroeconomic variables and commercial
property returns (listed and unlisted) in South Africa?
b) If such a relationship exists, what are the relationships between the chosen
macroeconomic variables and commercial property returns?
1.7 Research aim
This research aims to determine whether the relationship between macroeconomic variables
and commercial property returns, as found in international studies, is applicable to the South
African property environment.
9
1.8 Research proposition
This study will test the following research proposition:
There is a significant relationship between the performance of commercial property and
macroeconomic variables in South Africa.
1.9 Research objectives
This study aims to achieve the following research objectives:
a) establish the relevant economic variables;
b) determine whether there is significant linkage between economic variables and
commercial properties returns;
c) determine the extent of the relationship between economic variables and commercial
properties returns; and
d) assess the applicability of previous international studies.
1.10 Hypothesis
H0: There is no statistically significant relationship between macroeconomic variables and
commercial property (listed and unlisted) returns.
H1: There is a statistically significant relationship between macroeconomic variables and
commercial property (listed and unlisted) returns.
10
1.11 Methodology
This research builds on previous research providing the basis for testing both listed and unlisted
property returns (see Ling & Naranjo 1997; Hoesli, Lizieri & MacGregor 2008). The application
of vector autoregression (VAR) – employed by Downs et al. (2003), Payne (2003) and Laopodis
(2009) – provided the basis for the research methodology. This study undertook a literature
review on topics relevant to its field of study. The researcher then applied statistical methods
(correlation, multiple regression and vector autoregression) to selected data to formulate the
findings. Chapter 3 provides more details concerning the research methodology.
1.12 Justification of the research
Despite numerous international studies, no empirical research on the relationship between
economic variables and commercial properties returns has been conducted for South Africa.
The relationship between economic variables in this research is of significance to investors,
analysts and policymakers; such findings may assist them with investment and related decision-
making.
1.13 Limitations of this research
This study is subject to the following limitations:
a) The research focus is on commercial property only, due to data availability.
b) The research drew data from JSE and IPD, as the data are reliable and available to
public.
c) This research did not test all economic variables; certain variables were excluded due to
lack of available relevant data. The unavailability of data remains the major limitation to
this study; other South African studies have confirmed this limitation (see Franken,
Bloom & Erasmus 2011).
11
d) This research did not test firm specific variables – for example, gearing and market value
to book value.
e) This research focuses on the commercial property sector as a whole and not on the
influence of individual assets or companies.
f) The methodology adopted provides an understanding of historical experience and may
not be a predictive model.
g) The research periods selected were from 1995 to 2013 for unlisted properties and from
2002 to 2013 for listed properties.
h) Some of the research in the area identified the asymmetrical effects of interest rates to
asset prices (see Mueller & Pauley 1995; Simpson, Ramchander & Webb 2007; Chen et
al. 2012). Since the period of the research falls in a period of decreasing interest rates,
the results may be representative of a drop in interest rates and not throughout the
cycle.
1.14 Outline of the research report
Chapter 1 introduced the research background, problem, questions, aim, objectives,
hypothesis, methodologies and limitations. This concluding section presents an outline of the
remaining chapters of this thesis.
Chapter 2 presents a review of international and South African literature in this field. This
chapter addresses the question: ‘What has previous research in this field revealed?’
Chapter 3 draws together methodologies used and the economic variables identified in
previous research. It proposes the statistical research design to address the research questions.
Chapter 4 presents the results and interpretation of the statistical tests performed.
Chapter 5 concludes the discussion and recommends future areas of research.
This is followed by the full list of References and Appendices.
12
2. Literature Review
This chapter chronicles previous research undertaken in this field, focusing on both South
African and international literature.
Property is a financial asset and thus presumably is sensitive to economic variables, as financial
theory suggests that macroeconomic variables should systematically affect financial asset
returns (Chen, Roll & Ross 1986).
The relationship between financial assets returns and economic variables forms the basis of
finance theories (see section 2.1 for more details):
The capital asset pricing model (CAPM) considers co-movement of the market vis-à-vis
individual securities.
Arbitrage pricing theory (APT) considers macroeconomic variables as significant
variables in explaining financial asset returns.
The relationship between property returns and economic variables is well researched in
developed countries but has received less attention in South Africa, as most prior research in
South Africa focused on residential property. This identified gap presents a compelling research
opportunity.
Research of this nature is often accomplished through various statistical methods, including:
Correlation: This is a statistical method used to measure the strength or degree of
linear association between two variables (Gujarati 2003). This research uses this
method (see sections 3.1.1 and 3.1.2).
Multiple regression: This is a popular statistical technique frequently used in
econometrics and is concerned with the statistical dependence of dependent variable
on independent variable(s) (Gujarati 2003). This research uses this technique (see
section 3.1.1).
13
Vector autoregression: This technique is frequently used in econometrics for the
analysis of multivariate time series. Its main advantage is that researcher does not need
to assume structural inferences (Brooks & Tsolacos 1999). See section 3.1.2 for more
details.
Cointegration: Economically speaking, two variables will be cointegrated if they have a
long-term/equilibrium relationship (Gujarati 2003). This study did not apply this
methodology, as it is often applied as a pre-test to detect spurious regression.
To date, there is no scholarly literature that addresses this research area in South Africa. Thus,
this study and its findings are of significance to investors, analysts and policymakers.
Research by Chen, Roll and Ross (1986) provided the foundation for this area of research. They
based their research on the proposition that the CAPM provided for only one factor (the market
portfolio) as a determinant of financial markets and did not provide for ‘macroeconomic
variables’ that may impact financial markets (see Formula 2.1 in section 2.1). Chen, Roll and
Ross (1986) attempted to use macroeconomic variables to explain asset returns through the
context of APT, and used macroeconomic variables as variables in the APT return generation
process. The resulting empirical APT can be defined as the Macroeconomic Variable Model
(Chen, Hsieh & Jordan 1997).
Chen, Roll and Ross (1986) identified industrial production, changes in risk premium and the
term structure of interest rates2 as significant variables for explaining stock returns. They argue
that any economic variable that systematically affects either future cash flows and/or the
discount factor will impact the prices and returns of financial assets. Chen, Hsieh and Jordan
(1997) and Downs et al. (2003) support this view. They applied the methodology employed by
Chen, Roll and Ross (1986) that commenced the determinants of real estate returns.
DiPasquale and Wheaton (1992) set out a conceptual framework that divides the real estate
market into two underlying markets: the market for real estate space and the market for real
2 Term structure of interest rates is defined as the difference between long-term government yield less the Treasury bill rate.
14
estate assets. The variables identified in the conceptual framework included economy, rent,
supply, capitalisation rate, construction costs and replacements costs.
Ilmanen (2012, p. 112) identified variables such as economic growth, inflation, demographics
and population migration, as well as shorter-term supply-and-demand variables, as
fundamental determinants of property returns. However, Ilmanen (2012) also argues that
securitised properties, for instance, REITs are different in their behaviour and driven more by
equity markets and interest rates. He observed that fluctuations in cap rates and rental yields
are important drivers of real estate prices, often overwhelming the fundamental impact of
income growth.
The Appraisal Institute (2008, p. 44) points out: ‘To determine the influence of economic forces
on value, appraisers analyse the fundamental relationships between current and anticipated
supply and demand and the economic ability of the population to satisfy its wants, needs and
demands through its purchase power.’
Many specific market characteristics are considered in analyses of economic forces:
employment
wage levels
industrial expansion
the economic base of the region and the community
price level
the cost and availability of the mortgage credit
the stock of available vacant and improved properties
new development under construction or in the planning stage
occupancy rates
the rental and price patterns of existing properties
construction costs
Most prior studies have been on the securitised market, and the following variables were
identified to have significant impact on property returns:
15
Stock market (see Allen, Madura & Springer 2000; Okunev, Wilson & Zurbruegg 2000;
Payne 2003; He, Webb & Myer 2003; Standish et al. 2005; Clark & Daniel 2006; Huang &
Lee 2009; Franken, Bloom & Erasmus 2011; Chen et al. 2012; Yunus 2012).
Economic growth (see Chen, Roll & Ross 1986; Ling & Naranjo 1997; Ewing & Payne
2005; Clark & Daniel 2006; Franken, Bloom & Erasmus 2011; Boshoff & Cloete 2012;
Yunus 2012; Lieser & Groh 2013).
Interest rate (see Ling & Naranjo 1997; Allen, Madura & Springer 2000; Swanson, Theis
& Casey 2002; He, Webb & Myer 2003; Payne 2003; Clark & Daniel 2006; Huang & Lee
2009; Franken, Bloom & Erasmus 2011; Mangani 2011; Chen et al. 2012; Yunus 2012).
Inflation (see Adrangi, Chatrath & Raffliee 2004; Ewing & Payne 2005; Hoesli, Lizieri &
MacGregor 2008; Franken, Bloom & Erasmus 2011; Yunus 2012).
2.1 Stock market
The stock market is one of the most frequently used variables used to determine the
performance of listed property. Ilmanen (2012, p. 112) suggested that listed properties are
more driven by equity markets than market fundamental factors.
In terms of the CAPM theoretical framework, investors cannot diversify market risk away, but
they can diversify specific risk, and as an asset, market risk is the only risk that should be
rewarded (Hoesli & MacGregor 2000). Thus, the pricing of individual securities is expressed as a
co-movement of the market (Beta, see Equation 1, [2.1]).
16
Equation 1: Capital Asset Pricing Model
E (R) = RFR + β1 R(Mkt) [2.1]
Where:
E (R) Expected return of the portfolio
RFR Risk Free Rate
β1 Beta
R (Mkt) Expected return on the market
In terms of APT, Chen, Roll and Ross (1986) acknowledged the need to consider multiple risk
variables to price assets; however, a number of different variables have been identified (Hoesli
& MacGregor 2000).
Irrespective of the CAPM or APT model, the stock market index is often used as proxy to assess
market returns for research and valuation purposes. This thesis cites a number of studies that
used stock market returns as a proxy for market return (see Chen, Hsieh & Jordan 1997; Ling &
Naranjo 1997; Chen et al. 1998; Allen, Madura & Springer 2000; Swanson, Theis & Casey 2002;
Downs et al. 2003; He, Webb & Myer 2003; Simpson, Ramchander & Webb 2007; Huang & Lee
2009; Chen et al. 2012; Nittayagasetwat & Buranasiri 2012; Yunus 2012).
Most previous research has found positive relationships between listed property returns and
stock market returns (see Allen, Madura & Springer 2000; Okunev, Wilson & Zurbruegg 2000;
Payne 2003; He, Webb & Myer 2003; Huang & Lee 2009; Chen et al. 2012; Yunus 2012). These
observations are consistent with the study by Ilmanen (2012).
17
Liow (2010) studied 13 developed securitised real estate markets3 from 1989 to 2009. The
study determined that developed securitised real estate markets were more integrated with
their local stock market and weakly integrated with the global stock market and global real
estate markets over the past 20 years.
In South African literature, the ALSI is used as the proxy for market returns (see Standish et al.
2005; Clark & Daniel 2006; Franken, Bloom & Erasmus 2011).
According to a survey by PricewaterhouseCoopers (PwC 2012), the majority of corporate
finance practitioners in South Africa use CAPM as the pre-eminent model4 to determine cost of
equity (for corporate valuation) and use the All-Share Indices as the proxy for market portfolio.
2.2 Economic growth and the property market
According to the Fischer-DiPasquale-Wheaton (FDW) real estate model (DiPasquale & Wheaton
1992), economic growth translates into an increase in a number of variables in the property
market, thus affecting production, employment and household income (Du Toit & Cloete 2004).
Considering that these factors influence the demand for space, one assumes that they would
have a positive impact on the performance of the property market. Thus, in terms of economic
theory the relationship should be positive.
Other theoretical considerations include studies by Ilmanen (2012) and The Appraisal Institute
(2008), who considered economic growth as fundamental determinants of property returns.
The results from previous research on economic growth and property prices/returns were
mixed. Most studies found significant positive relationships (Chen, Roll & Ross 1986; Ling &
Naranjo 1997; Clark & Daniel 2006; Franken, Bloom & Erasmus 2011; Boshoff & Cloete 2012;
Yunus 2012; Lieser & Groh 2013), which is consistent with economic theory. Some studies
3 The author selected the countries as they were included in the S&P Global Property Index, namely: Australia, Hong Kong, Japan, Singapore, Belgium, France, Netherlands, Spain, Sweden, Switzerland, the United Kingdom, Canada and the United States. 4 The other methods used included APT, dividend growth models and others.
18
found negative relationships (Ewing & Payne 2005), and others found that economic variables
have no significant interaction with property returns (Brooks & Tsolacos 1999; Standish et al.
2005; Chang, Chen & Leung 2011).
The inconsistency in influence of the variables on property returns can be attributed to
different periods, data sets, different types of proxies for economic growth and variety of
methodologies used (see Table 3).
Table 4 summarises the relationships between economic growth and property returns.
Table 3: Summary of literature review (showing inconsistency between period, frequency of data, methodologies and results )
Studies Country, period of analysis, frequency of data
Data used (independent variable)
Methodologies Results between economic growth and independent variable
Chen, Roll & Ross (1986)
US, 1953 – 1983, monthly
NYSE listed stocks
Regression Positive
Ling & Naranjo (1997)
US, 1978 – 1994, Quarterly
REIT and NCREIF5 data
Regression Positive
Brooks & Tsolacos (1999)
UK, 1985 – 1998, Monthly
REIT VAR (Variance decomposition and Impulse Response)
No/inconclusive
Ewing & Payne (2005)
US, 1980 – 2000, Monthly
REIT VAR (Impulse Response)
Negative
Standish et al. (2005)
SA, 1974 – 2003, Quarterly
ABSA Housing Index
Regression No/inconclusive
Clark & Daniel (2006)
SA, 1980 – 2006, Quarterly
Residential house price
Regression Positive
Chang, Chen & Leung (2011).
US, 1975 – 2008, Quarterly
REIT VAR (Variance decomposition and Impulse Response)
No/inconclusive
Franken, Bloom & Erasmus
SA, 1974 – 2004, Quarterly
Residential house price
Regression Positive
5 The National Council of Real Estate Investment Fiduciaries (NCREIF) is a not-for-profit trade association that provides its members with commercial properties data, performance measurement and investment analysis. (http://www.ncreif.org)
19
(2011)
Boshoff & Cloete (2012)
SA, 2000 – 2009, not disclosed
Used Share Price of PLS
Correlation Positive
Yunus (2012) US, Canada, Japan, Australia, Germany, France, Italy, Netherlands, Switzerland and UK, 1990 – 2007, Monthly and Quarterly
REIT Cointegration and VAR (Granger Causality and Impulse Response)
Positive
Lieser & Groh (2013)
47 countries, 2000–2009, Annual
Commercial Real Estate Investment in USD million
Augmented Panel Regression Analysis
Positive
Table 4: Summary of relationships between economic growth and property returns
Relationship Studies
Positive Chen, Roll & Ross (1986); Ling & Naranjo (1997); Clark & Daniel (2006);
Franken, Bloom & Erasmus (2011); Boshoff & Cloete (2012); Yunus (2012);
Lieser & Groh (2013)
Negative Ewing & Payne (2005)
No/inconclusive Brooks & Tsolacos (1999); Standish et al. (2005); Chang, Chen & Leung
(2011)
20
2.3 Interest rates
It is often assumed that property returns are linked to interest rates, because changes in
interest rate impact return on property investment in two ways:
1. They impact the cash return of property, as most of the property investments have
some form of financial gearing and interest rate changes will impact their bottom line.
(Lynn 2007).
2. They impact the discount rate/capitalisation rate used to value the property
investments (Chen, Roll & Ross 1986; Ilmanen 2012).
In terms of the FDW real estate model (DiPasquale & Wheaton 1992), ‘The demand for real
estate assets is determined by real estate yields in relation to the after tax yield of fixed income
securities and other investments’ (Du Toit & Cloete 2004). If the interest rate in the rest of the
economy rises, then the existing ‘yield’ from the real estate becomes too low relative to fixed-
income securities and investors will wish to shift their funds from the property sector. Thus the
capitalisation rate will rise and depress property prices.
The International Valuation Standards Framework, paragraph 60, outlines the ‘Income
Approach’ valuation methodology: ‘This approach considers the income that an asset will
generate over its useful life and indicates value through a capitalisation process. Capitalisation
involves the conversion of income into a capital sum through the application of an appropriate
discount rate’ (International Valuation Standard Council 2011, p. 26). This method is consistent
with the ‘Income Capitalisation Approach’ advanced by The Appraisal Institute in which ‘a
property’s income and resale value upon reversion may be capitalised into a current, lump sum
value’ (The Appraisal Institute 2008, p. 142).
Therefore the theoretical argument is that the influence between property returns and the
interest rate is negative, since an increase in the interest rate will increase the
discount/capitalisation rate, or an increase in the interest rate will increase the interest
repayments and reduce the bottom line of the property. Either way, an increase in the interest
rate will reduce the value of the investment and returns on the property.
21
Past research has produced mixed results. The majority of prior studies have found significant
negative relationships between property returns and the interest rate (see Allen, Madura &
Springer 2000; He, Webb & Myer 2003; Payne 2003; Clark & Daniel 2006; Huang & Lee 2009;
Mangani 2011; Chen et al. 2012; Yunus 2012). This is consistent with theoretical arguments.
However, Chen et al. (2012) applied quantile regression testing, which further indicates that the
impact of monetary policy (Fed Rate is used as a proxy) has a differential impact on the REIT
market. The impact is significantly negative during bull markets and has no impact during
bear/volatile markets.
Some prior studies have found significant positive relationships (see Ling & Naranjo 1997;
Swanson, Theis & Casey 2002; Franken, Bloom & Erasmus 2011). This relationship between
interest rates and equity REITs can be explained as, despite the proposition that lower interest
rates reflect weak economic conditions and low inflationary expectations, an increase in
interest rates may reflect stronger economic growth, higher inflationary expectations, and
upward pressure on real estate prices. These effects may result in a positive relationship or
negate the inverse relationship between interest-rate movements and real estate values. (See
2.3.1 for further discussion and empirical research.)
Research has also found that the interest rate has little impact on the performance of property
returns (see Mueller & Pauley 1995; Chen et al. 1998; Nittayagasetwat & Buranasiri 2012).6
The inconsistency in influence of interest rate on property returns can be attributed to different
periods, data sets and different proxies used for interest rates (see 2.3.1) deployed in prior
studies (see Table 5).
6 Most interest proxies not significant, with the exception of long-term US high-grade corporate bonds.
22
Table 5: Summary of l iterature review ed above (showing inconsistency between period, frequency of data, methodologies and results
Studies Country, period
of analysis,
frequency of data
Data used
(independent
variable)
Methodologies Results between
interest rate and
independent
variable
Mueller & Pauley
(1995)
US, 1972 – 1993,
monthly
REIT Regression No relationship
Ling & Naranjo
(1997)
US, 1978 – 1994,
Quarterly
REIT and
NCREIF7 data
Regression Significant positive
relationship to
Treasury bill rate
Chen et al. (1998) US, 1978 – 1994,
Monthly
REIT Regression No relationship
Swanson, Theis &
Casey (2002)
US, 1989 – 1998,
Daily
REIT Regression Risk premium of 30
year Treasury bond
has significant
positive relationship
REIT return
Franken, Bloom &
Erasmus (2011)
SA, 1974 – 2004,
Quarterly
Residential
house price
Regression Interest rate has
positive relationship
to residential
property prices
Nittayagasetwat &
Buranasiri (2012)
US, 2000 – 2011,
Monthly
REIT Regression The monthly return
on long-term US
high-grade
corporate bonds was
the statistically
significant interest
rate proxy, which
affected REIT’s
performance. All
other interest
proxies not
significant.
7 National Council of Real Estate Investment Fiduciaries (NCREIF) is a not-for-profit trade association that provides its members with commercial properties data, performance measurement and investment analysis. http://www.ncreif.org
23
2.3.1 Types of interest rate proxies
Previous research has suggested a number of proxies for the impact of interest rates:
Brooks and Tsolacos (1999) and Huang and Lee (2009) used the nominal interest rate,
defined as ‘real interest rate plus a premium for expected inflation’ (The Appraisal
Institute 2008, p. 97).
Other studies used the real interest rate, defined as the nominal Treasury bill rate less
inflation (see Chen, Roll and Ross 1986; Standish et al. 2005).
Other studies used the risk premium, defined as the Baa bond8 yield less long-term
government yield (see Chen, Roll & Ross 1986; Chen, Hsieh & Jordan 1997; Chen et al.
1998; Swanson, Theis & Casey 2002; He, Webb & Myer 2003; Payne 2003; Ewing &
Payne 2005; Nittayagasetwat & Buranasiri 2012).
Still other studies used the term structure, defined as the difference between long-term
government yield less the Treasury bill rate (see Chen, Roll & Ross 1986; Ling & Naranjo
1997; Chen, Hsieh & Jordan 1997; Chen et al. 1998; Swanson, Theis & Casey 2002; He,
Webb & Myer 2003; Payne 2003; Nittayagasetwat & Buranasiri 2012).
He, Webb & Myer (2003) found that overall the changes in yields on high-yield corporate bonds
(Baa) have the strongest explanatory power for returns on REITs for most of the 27-year sample
period (1972 – 1998).
Short-term vs long-term interest rates
South African proxies used for short-term interest rates are the South African 3-month Treasury
bill rate (see Das, Gupta, Kanda, Tipoy & Zerihyn 2012) and the negotiable certificate of deposit
(see Hassan & Biljon 2009).
8 Bond rating by Moody’s
24
Long-term interest rate proxies include: JSE Actuaries All-Bond Index used by Hassan and Biljon,
(2009), the R1869 bond used by majority of corporate financiers (PwC 2012) and the prime
interest rate (Clark & Daniel 2006; Franken, Bloom & Erasmus 2011).
Investments rate vs borrowing rate
In the South African context, two major proxies are used as a benchmark for interest rates. The
first is the 10-year government bond rate, which is considered as a risk-free interest rate
benchmark, as it is the rate investors can obtain by investing in a long-term South African
government bond. The other is the prime overdraft rate set by the Reserve Bank of South
Africa, commonly known as the Prime Interest Rate. Most commercial bank interest rates in
South Africa are linked to the Prime Interest Rate.
Listed properties are an asset class that is often compared to government bonds, since ‘the long
run return of real estate is between that of bonds and stocks’ (Ilmanen 2012, p. 102). This can
be observed from the opinions of numerous investment analysts and asset managers that listed
properties are highly correlated to long-term government bonds or used as a benchmark to
determine the attractiveness of listed property as an asset class (see Appendix 1).
According to PwC (2012), the proxies used by business valuation practitioners for the risk-free
rate are predominantly South African government bonds (79% of respondents; in the 2010
Survey, 93% of the respondents used South African government bonds).
Based on the analysis of this thesis (see Figure 2), there is a high level of association between
the ten-year yield and the Prime Interest Rate. The correlation between the ten-year yield and
the Prime Interest Rate is 89% for the period from 1960 to 2013. Thus using either the ten-year
yield or the Prime Interest Rate as proxy for the interest rate should yield similar results.
9 R186 is a bond issued by South African government with a maturity date of 11/12/2026.
25
Figure 2: Investment rate vs borrowing rate (Bureau of Economic Research, University of
Stellenbosch, prepared by author)
2.3.2 Non-academic empirical studies
Non-academic empirical studies performed by asset managers and consultants have produced
various results.
Towers Watson (2012) found no relationship between gilt yield and property returns.
Blackrock Real Estate Equity Group (2005), CBRE Clarion Securities (2013), and JP Morgan Asset
Management (2013) found that in a rising interest rate environment, returns actually increase,
since rising interest rates are normally consistent with improving economic conditions.
Blackrock (2005) found that returns (income and capital) of commercial properties increase
during periods of rising interest rates relative to periods when interest rates decline.
CBRE (2013) found that in the period of a rising interest rate in which REIT corrected more than
10%, REITs total returns underperformed equity market returns in the short term but thereafter
delivered subsequent periods of strong absolute returns, and generally outperformed the
broader equity market.
0%
5%
10%
15%
20%
25%
30%
Jan
-60
Sep
-61
May
-63
Jan
-65
Sep
-66
May
-68
Jan
-70
Sep
-71
May
-73
Jan
-75
Sep
-76
May
-78
Jan
-80
Sep
-81
May
-83
Jan
-85
Sep
-86
May
-88
Jan
-90
Sep
-91
May
-93
Jan
-95
Sep
-96
May
-98
Jan
-00
Sep
-01
May
-03
Jan
-05
Sep
-06
May
-08
Jan
-10
Sep
-11
May
-13
10 Year Yield Prime Interest Rate
26
JP Morgan (2013) found that in the short term, rising interest rates have a limited impact on
capitalisation rates, as a rising interest rate is normally consistent with improving economic
conditions. In the long term, the current core unleveraged yield on property is the main
determinant of whether rising interest rates will impact property prices.
Cohen and Steers Capital Management (2013) found that during periods of rising inflation, REIT
returns outperformed equity markets. Capitalisation rates (cap rates) also do not move in
tandem with interest rates but with economic growth expectations and credit spreads.
Table 6 summarises the above non-academic research by country, period of analysis, used data,
methodologies, and the relationship between property returns and the interest rate.
27
Table 6: Summary of l iterature review (non-academic research)
Author Country, period
of analysis,
frequency of
data
Data Used
(independent
variable)
Methodologies Relationship between property
returns and the interest rate
Towers
Watson
(2012)
UK 1987 – 2010,
Annually
IPD Correlation
Scatterplot
No significant relationship.
Blackrock
(2005)
US 1978 – 2004,
Quarterly
NCREIF Found returns income and
capital) of commercial properties
increase during period of rising
interest rates.
CBRE (2013) US 1994 – 2013,
Daily
REIT Event study and
charts
In the period of rising interest
rate REIT corrected more than
10%. In the short term, REIT total
returns underperformed equity
market returns, thereafter,
delivered subsequent periods of
strong absolute returns, and
generally outperformed equity
market.
JP Morgan
(2013)
US, 1983 –
2013, Quarterly
REIT Line chart and
Scatter plot
In the short term, rising interest
rates have limited impact on
capitalisation rates.
Cohen &
Steers
(2013)
US 1979 – 2012,
Monthly
REIT Event study and
charts
During period of rising inflation,
REIT returns outperformed equity
markets, capitalisation rate do
not move in tandem with interest
rates and that REITs can be an
effective inflation hedge
28
2.4 Inflation
The Proxy Effect Hypothesis, initially developed by Fama (1965 cited in Adrangi, Chatrath &
Raffliee 2004), describes the negative relationship between equity returns and inflation. The
hypothesis predicts that rising inflation rates reduce real economic activity and demand for
money. The decline in economic activity should subsequently negatively affect employment and
stock returns (Adrangi, Chatrath & Raffliee 2004).
Non-academic empirical research conducted by Cohen and Steers (2013) found the contrary:
US REITs can be effective as a hedge against inflation, since US REITs have outperformed stocks
and bonds in periods of both rising and moderating inflation through dividend growth at a pace
faster than inflation.
The relationship between asset returns and inflation has been extensively researched,
particularly in terms of the effectiveness of REIT in hedging inflation. The studies are often
divided into two types of proxies – actual inflation and expected/unexpected inflation.
Yobaccio, Rubens and Ketcham (1995) studied the inflation-hedging property of REIT from 1972
to 1992 and found that REIT acts as poor hedge against any measure of inflation actual,
expected or unexpected).
2.4.1 Actual inflation
The proxy predominantly used by researchers for actual inflation is the consumer price index
(CPI) (see Chen, Roll & Ross 1986; Yobaccio, Rubens & Ketcham 1995; Chen, Hsieh & Jordan
1997; Chatrath & Liang 1998; Chen et al. 1998; Glascock, Lu & So 2002; Payne 2003; Adrangi,
Chatrath & Raffliee 2004; Ewing & Payne 2005; Franken, Bloom & Erasmus 2011; Yunus 2012).
The results of the research are not consistent, with the majority of the studies finding no
relationship/response or a negative relationship/response.
29
Despite the different time periods, most studies found no relationship (see Chen, Hsieh &
Jordan 1997; Chatrath & Liang 1998; Chen et al. 1998; Glascock, Lu & So10 2002).
Some studies found a negative relationship (see Adrangi, Chatrath & Raffliee 2004; Ewing &
Payne 2005). However, a few studies found a positive relationship/response (see Franken,
Bloom & Erasmus 2011; Yunus 2012).
Table 7: Summary of l iterature reviews above (showing inconsistency between period, frequency of data, methodologies and results
Studies Country, period of analysis, frequency of data
Data used (independent variable)
Methodologies Results between actual inflation and independent variable
Chen, Hsieh & Jordan (1997)
US, 1974 – 1991, Monthly
REIT Regression None
Chatrath & Liang (1998)
US, 1972 –1995, Monthly
REIT Regression and Cointegration tests
No relationship between inflation and REIT.
Chen et al. (1998) US, 1978 – 1994, Monthly
REIT Regression None
Glascock, Lu & So (2002)
US, 1972 – 1995, Monthly
REIT VECM, VAR (Variance decomposition and Impulse Response)
REIT returns anticipate changes in inflation (expected and unexpected).
Adrangi, Chatrath & Raffliee (2004)
US, 1972 – 1999, Monthly
REIT Regression and Cointegration tests
Real REIT returns are negatively correlated with inflation.
Ewing & Payne (2005)
US, 1980 – 2000, Monthly
REIT VAR (Generalised impulse response)
Shocks to monetary policy lead to lower than expected returns
Chang, Chen & Leung (2011)
SA, 1974 – 2004, Quarterly
Residential house price
Regression Inflation has positive relationship to residential house price.
Yunus (2012) US, Canada, Japan, REIT Cointegration and ‘shocks to inflation
10 Glascock, Lu and So (2002) concluded that the observed negative relationship between REITs and inflation is spurious, and this is explained once the monetary policy effects on the respective variables are specifically taken into account.
30
Australia, Germany, France, Italy, Netherlands, Switzerland and UK, 1990 – 2007, Monthly and Quarterly
VAR (Granger Causality and Impulse Response)
induce a positive response in international securitised property returns’
2.4.2 Expected/unexpected inflation
Expected and unexpected inflation are derived and calculated variables:
Expected inflation is defined as the Treasury bill rate less ex post real rate of interest
(see Fama and Gibbons 1984 cited in Chen, Roll and Ross 1986).
Unexpected inflation is defined as the difference between the realised inflation rate
during period t and the expected inflation rate at the beginning of the same period t.
The realised inflation rate is the first-order log relative of the CPI for all urban
consumers. Unexpected inflation is calculated by the Fama and Gibbons (1984 cited in
Chen, Roll and Ross 1986) method (see Equation 2, [2.2]), which uses the Fisher
equation and time-series analysis to derive unexpected inflation.
Equation 2: Fisher theorem on deriving unexpected inflation
UI(t) = I(t) – E[I(t)│t–1] [2.2]
UI(t) is unexpected inflation.
I(t) is the realised monthly first difference in the logarithm of CPI for period t.
E[I(t)│t–1] is expected inflation defined as Treasury bill rate less ex post real
rate of interest.
Most studies found expected inflation insignificant by (see Chen, Hsieh & Jordan 1997; Ling &
Naranjo 1997; Chen et al. 1998).
31
Hoesli, Lizieri and MacGregor (2008) is the only study that found expected inflation positively
linked to asset returns.
Most researchers found no significant relationship between unexpected inflation and real
estate returns (see Ling & Naranjo 1997; Chen, Hsieh & Jordan 199711; Chen et al. 1998; Brooks
& Tsolacos 1999).
However, Simpson, Ramchander and Webb (2007) found previous studies to be flawed. Their
study documented an asymmetrical response of REIT return to inflation. During expansionary
periods, REIT returns go up with both increases and decreases in inflation. However, during a
restrictive monetary policy period, the asymmetric framework cannot explain the perverse
relationship between REIT return and inflation.
2.5 Summary of the literature review
Table 8 summarises the relevant international academic literature by country, data used,
methodologies and variables found to be significant in the research. The table is sorted in the
order of the research studies quoted in the literature review.
Table 8: Summary of academic literature review (in alphabetical order)
Author Country, period of analysis, frequency of data
Data Used
Methodologies Variables found to be significant
Adrangi, Chatrath & Raffliee (2004)
US, 1972 – 1999, Monthly
REIT Regression and Cointegration tests
Real REIT returns are negatively correlated with inflation.
Allen, Madura & Springer (2000)
US, 1993 –1997, Monthly
REIT Regression REITs are sensitive to long- or short-term interest-rates.
Boudry, Coulson,
US, 1984 – 2009,
REIT VECM REITs and the underlying real estate markets are related. Furthermore, the
11 for two of the three periods tested (January 1980 to December 1985 and January 1986 to December 1991)
32
Kallberg & Liu (2012)
Quarterly and Annually
relation appears to be stronger in particular in annual rather than quarterly data.
Brooks & Tsolacos (1999)
UK, 1985 –1998, Monthly
REIT VAR (Variance decomposition and Impulse Response)
The conclusion from the VAR methodology adopted in this paper is that the overall, UK real estate returns purged of general stock market influences are difficult to explain on the basis of the information contained in the macroeconomic variables tested.
Chatrath & Liang (1998)
US, 1972 –1995, Monthly
REIT Regression and Cointegration tests
This study concludes that there is no relationship between inflation and REIT.
Chen (1997) US, 1974 – 1991, Monthly
REIT Regression None
Chen et al. (1998)
US, 1978 – 1994, Monthly
REIT Regression None
Chen et al. (2012)
US, 1972 – 2008, Monthly
REIT Regression The results of this study show that stock market returns had a positive impact on EREIT returns in the period from 1972 to 2008.
Downs et al. (2003)
US, 1972 – 1999, Monthly
REIT VAR (Variance decomposition and Impulse Response)
The relationship between income-return variance and each variable is statistically significant; construction accounts for about 10%; industrial production, 4%; T-bill yield, 5%; mortgage rate, 7%; and market portfolios, 2%. Only the past price return is consistently significant in price-return variance.
Ewing & Payne (2005)
US, 1980 – 2000, Monthly
REIT VAR (Generalised impulse response)
The research examined data from 1980 to 2000. This research found that shocks to monetary policy, economic growth and inflation all lead to lower than expected returns, while a shock to default risk premium is associated with higher return.
Glascock, Lu & So (2002)
US, 1972 – 1995,
REIT VECM, VAR (Variance
Fed Fund Rate provided partial explanation to the relationship
33
Monthly decomposition and Impulse Response)
between REIT returns and REIT returns anticipate changes in inflation (expected and unexpected).
Huang & Lee (2009)
US, 1994 – 2007, Daily
REIT Autoregressive Integrated Moving Average (ARIMA)
Huang and Lee (2009) investigated asset returns from 1994 to 2007. They found the changes the REIT is negatively sensitive to interest rates and demonstrate that REIT returns are more sensitive in the long term than in the short term.
He, Webb & Myer (2003)
US, 1972 – 1998, Monthly
REIT Flexible Least Square
This study found changes in yields on high-yield corporate bonds (Baa).
Hoesli, Lizieri & MacGregor (2008)
US and UK, 1977 – 2003, Quarterly
REIT, NCREIF (US) and IPD (UK)
Error Correction Model
Hoesli, Lizieri & MacGregor (2008) found that anticipated inflation was positively linked to asset return. The ECM approach clearly demonstrates that asset return adjustment to changes in inflation does not occur in period but rather through an error correcting adjustment process to the long run relationship that is gradual. The results were similar between the UK and the US.
Laopodis (2009)
US, 1971 – 2007, Monthly
REIT VAR (Variance decomposition and Impulse Response)
Laopodis (2009) found that REITs display reciprocal linkage between the general stock market and industrial production movements.
Lieser & Groh (2013)
47 countries, 2000–2009, Annual
Commercial Real Estate Investment (USD million)
Augmented Panel Regression Analysis
GDP per capita and inflation
Ling & Naranjo (1997)
US, 1978 – 1994, Quarterly
REIT and NCREIF12 data
Regression Real per capita consumption growth
Liow (2010) 13 developed countries,
REIT Cointegration test
The developed securitised real estate markets are more integrated with their
12 National Council of Real Estate Investment Fiduciaries (NCREIF) is a not-for-profit trade association that provides its members with commercial properties data, performance measurement and investment analysis. http://www.ncreif.org
34
1989 – 2009, Weekly
local stock market.
Mueller & Pauley (1995)
US, 1972 – 1993, Monthly
REIT Regression The result indicated that the price movement has a low correlation with changes in interest rate and a lower correlation of interest rate than with movement in the stock market as a whole.
Nittayagasetwat & Buranasiri (2012)
US, 2000 – 2011, Monthly
REIT Regression The monthly return on long-term US high-grade corporate bonds was the statistically significant interest rate proxy, which affected REIT performance.
Okunev, Wilson & Zurbruegg (2000)
US, 1972 – 1998, Monthly
REIT VAR Granger (Causality)
Strong unidirectional non-linear relationship running from the stock market to REIT was found.
Payne (2003) US, 1982 –2003, Monthly
REIT VAR (Impulse Response)
Unexpected shocks in the broad stock market index have a positive impact on REITs. An unexpected shock to the term structure has an adverse effect on REITs.
Simpson, Ramchander & Webb (2007)
US, 1981 – 2002 and 1990 – 2002, Monthly
REIT Regression During expansionary periods, REIT returns go up with both increases and decreases in inflation. However, during restrictive monetary policy periods, the asymmetric framework cannot explain the perverse relationship between REIT returns and inflation.
Swanson, Theis & Casey (2002)
US,1989 – 1998, Daily
REIT Regression They found that value weighted stock index and risk premium of a 30-year Treasury bond appears to explain REIT return better than other interest rate proxy used.
Yunus (2012) US, Canada, Japan, Australia, Germany, France, Italy, Netherlands, Switzerland and U.K. 1990 – 2007,
REIT Cointegration and VAR (Granger Causality and Impulse Response)
Property markets co-integrated with its respective stock markets in the long run. In the majority of the countries investigated, property returns positively responded to shocks to the stock market.
35
Monthly and Quarterly
2.6 Other variables
Previous studies have identified other economic variables that are not considered in this study:
demographics (Lynn 2007)
employment/unemployment levels (Brooks & Tsolacos 1999)
the cost and availability of mortgage credit (Downs et al. 2003)
new development under construction or in the planning stage (Clark & Daniel 2006;
Franken, Bloom & Erasmus 2011)
construction costs (Downs et al. 2003; Franken, Bloom & Erasmus 2011)
oil (Chen, Roll & Ross 1986; Clark & Daniel 2006; Hoesli, Lizieri & MacGregor 2008,
Huang & Lee 2009)
motor vehicle sales (Clark & Daniel 2006)
debt/household income (Standish et al. 2005; Clark & Daniel 2006; Franken, Bloom &
Erasmus 2011)
liquidity (Ilmanen 2012)
dividend yield (Downs et al. 2003; Brooks & Tsolacos 1999)
The variables above were not considered for this study as they are not frequently tested, lack
reliable data, were not necessary economic variables and were not found to be significant
variables in prior studies.
36
2.7 South African literature
Currently, research into the South African property market is limited, with the majority of the
focus placed on performance in the residential sector.
Standish et al. (2005) focused on isolating the determinants of residential property in South
Africa and developed two national models. One model covers the period from 1974 to 2003
and another model covers the period from 1994 to 2003. For the 1974 to 2003 model,
significant variables were net immigration (positive relationship), real capitalisation of JSE
(negative relationship), foreign direct investment (positive relationship), the real Rand gold
price (positive relationship) and the Rand dollar exchange rate (negative relationship). For the
1994 to 2003, significant variables were the ratio of household debt to income (negative
relationship), foreign direct investment (positive relationship) and the real Rand gold price
(positive relationship).
Clark and Daniel (2006) developed a forecast model for South Africa’s residential housing
market. They identified the following variables to forecast South African residential housing
prices: All-Share Index (positive relationship), GDP (positive relationship), Prime Interest Rate
(negative relationship), Rand/US Dollar exchange rates (negative relationship) and transfer
costs (positive relationship).
In their research, Franken, Bloom and Erasmus (2011) identified eight indicators that could be
utilised as predictors of future residential estate price cycles. The variables were: construction
costs, consumption, the debt to income ratio, GDP, inflation, interest rate, the JSE ALSI and
affordability. All the variables are positively related to residential properties prices.
Simo-Kengne, Bittencourt and Gupta (2012) investigated the economic impact of house prices
in South Africa using a panel data set that covered all nine provinces from 1996 to 2010. They
found strong evidence that economic growth affects house prices.
Further, a working paper by Simo-Kengne, Bittencourt and Gupta (2013) found that in South
Africa house price changes exhibit a significant effect on regional economic growth. The paper
37
applied a VAR model to investigate the extent to which macroeconomic shocks are responsible
for the common component in house price movements. The results indicate that all macro
shocks have significant influences on real house prices with portfolio shocks having the largest
fraction in the total variability in real house prices followed by monetary policy shocks. This
finding substantiates the user-cost theory, which emphasises the importance of interest rates
and expectations in driving house price dynamics. Thus, there is evidence to suggest that during
periods of high volatility in house prices, interest rates fall steadily and people expect strong
growth in house prices, resulting in lower user costs of housing, which in turn increases
property prices.
Du Toit and Cloete (2004) provided one of few commercial property studies that considered the
development of an integrated property and asset market model for South African property
markets, utilising the Pretoria office market as case study. They apply the FDW real estate
model (DiPasquale & Wheaton 1992) to simulate the interrelationships between property and
asset markets in a diagrammatic quadrant model configuration.
Mangani (2011) applied the Generalised Autoregressive Conditional Heteroskedasticity
(GARCH) model to investigate monetary policy on the JSE portfolio for the period from 1990 to
2009. The analysis showed that repo rates changes are important for describing mean return
and return volatilities. The repo rate has a significant negative coefficient; this indicates that
contractionary monetary policy tended to lower stock returns as theoretically postulated.
However, the effect of repo rate changes were found to be asymmetrical, i.e., the impact of the
repo rate changes differs for positive rate changes versus a negative rate changes. The results
suggest that JSE returns are more responsive to contractionary monetary policy than to
expansionary policy
Boshoff and Cloete (2012) applied a correlation to determine the relationship between
property share price and economic variables. They found that the share price of PLS is
correlated to employment (in private, public and non-agriculture sector), disposal income,
national government revenue and expenditure, GDP (at market prices) and gross value added
at basic prices of construction (a measure of construction industry activity used in calculation of
38
GDP) and repo rates. However, they pointed out that there are limitations to their research, as
they applied a simple linear correlation and thus excluded the combined effect of more than
one variable.
Table 9 summarises South African academic literature reviewed by period, frequency, data
used, methodologies and variables found to be of significance.
39
Table 9: Summary of South African academic l iterature review
Author Country, period
of analysis,
frequency of
data
Data Used Methodologies Variables found to be significant
Standish
et al.
(2005)
SA, 1974 –
2003, Quarterly
ABSA
Housing
Index
Regression They found that net immigration
(positive relationship), real
capitalisation of JSE (negative
relationship), foreign direct
investment (positive relationship),
real Rand gold price (positive
relationship), and rand dollar
exchange rate (negative
relationship) to be significant
variables.
Clark &
Daniel
(2006)
SA, 1980 –
2006, Quarterly
Residential
house price
Regression All-Share Index (positive
relationship), GDP (positive
relationship), Prime Interest Rate
(negative relationship), Rand/US
Dollar exchange rates (negative
relationship), and transfer costs
(positive relationship)
Franken,
Bloom &
Erasmus
(2011)
SA, 1974 –
2004, Quarterly
Residential
house price
Regression constructions cost, consumption,
the debt to income ratio, GDP,
inflation, interest rate, the JSE
ALSI, and affordability (all positive
relationship)
Boshoff
and Cloete
(2012)
SA, 2000 –
2009, not
disclosed
Used Share
Price of PLS
Correlation Repo interest rate (negative
relationship)
GDP (positive relationship)
40
2.8 Summary and conclusion
This chapter chronicles existing literature in this field both in South Africa and internationally
and is summarised as follows:
Property is a financial asset, and as financial theory suggests should be sensitive to economic
variables (Chen, Roll & Ross 1986). Theoretical frameworks such as DiPasquale and Wheaton
(1992), Ilmanen (2012) and prior empirical research summarised earlier have identified the
following variables as being of critical importance:
The performance of the stock market (section 2.1) is considered as one of the most
frequently used variables in prior research, and the majority of the research found a
positive relationship between property returns and stock market performance.
Economic growth (section 2.2) was identified in the various theoretical frameworks
(DiPasquale & Wheaton 1992; The Appraisal Institute 2008); prior empirical research
yielded mixed results. (The majority of studies found a positive relationship, which is
consistent with economic theory.)
Interest rates (section 2.3) influence property returns through the impact on cash return
of the property and/or capitalisation rate. The results of empirical research are mixed,
but the majority found a negative relationship between property returns and the
interest rate; this negative relationship is consistent with theoretical arguments. Non-
academic research found a mostly positive relationship between property returns and
interest rates. The inconsistency can be attributed to different period, data, proxies13
and methodologies deployed.
Inflation (section 2.4) is an extensively researched variable. The often used proxies are
actual inflation (section 2.4.1), expected inflation (section 2.4.2) and unexpected
inflation (section 2.4.2). Most past studies found no significant relationship between
inflation (actual, expected and unexpected) and property returns.
13 See section 2.3.1: nominal nterest rate, real interest rate, risk premium, term structure, change in expected inflation, short-term interest rates, long-term interest rates, investment rate, and borrowing rate
41
South African literature (section 2.7) is limited, with the majority of studies focused on
residential house price. The stock market and GDP were found to have a positive
relationship (Clark & Daniel 2006; Franken, Bloom & Erasmus 2011; Boshoff & Cloete
2012) to residential property prices. The interest rate was found to have both positive
(Clark & Daniel 2006) and negative relationships (Franken, Bloom & Erasmus 2011;
Boshoff & Cloete 2012).
Chapter 4 will test the selected variables.
42
3. Research Methodology
This research is based on a deductive approach and brings together the theoretical framework,
relevant statistical methodologies and economic variables identified in previous studies as
presented in Chapter 2. This chapter provides further details on how proxies for the chosen
economic variables were selected and how the data was collected.
Based on economic theories and a review of previous international and local studies, it seems
reasonable to suggest that there are statistically significant economic variables that influence
commercial property returns.
3.1 Statistical methodology
The objective of this study is to investigate and identify the macroeconomic variables that
systematically affect commercial properties returns. There are two methodologies applied in
this study, namely: cross-sectional regression and vector autoregression (VAR). These two
statistical approaches were the most frequently used14 in prior research and were chosen for
this research. As discussed in the previous chapters, previous research applied either regression
or VAR models. This study applies both regression and VAR methodologies.
In order to assess the robustness of this research, this study considered both listed and unlisted
commercial properties. IPD data is the proxy used for listed commercial properties are listed
property indices and proxy for unlisted commercial properties. This approach of testing both
listed and unlisted data is similar to the approach of prior studies (see Ling & Naranjo 1997;
Hoesli, Lizieri & MacGregor 2008). Table 10 summarises the statistical methodologies applied
for this research.
14 Seventeen prior literatures cited in this research used regression (see 3.1.1 for the references) and ten prior literatures used VAR (see section 3.1.2 for details).
43
Table 10: Summary of statistical methodologies to be applied
Type of commercial
properties
Methodologies
Unlisted Cross-sectional regression
Unlisted VAR, variance decomposition and impulse response
Listed Cross-sectional regression
Listed VAR, variance decomposition and impulse response
Most prior research focused narrowly on one or two macroeconomic variables only and applied
either one of the statistical methodologies mentioned above (see Table 11).
Table 11: Summary of academic literature review (in alphabetical order)
Author Methodologies Variables considered in the research
Adrangi, Chatrath & Raffliee (2004)
Regression and cointegration tests Inflation
Allen, Madura & Springer (2000)
Regression Interest rate and stock market
Boudry, Coulson, Kallberg & Liu (2012)
VECM Stock market (various proxies)
Brooks & Tsolacos (1999)
VAR (variance decomposition and impulse response)
Rate of unemployment, interest rates (various proxies), inflation and the dividend yield
Chatrath & Liang (1998) Regression and cointegration tests Inflation
Chen, Hsieh & Jordan (1997)
Regression Inflation (various proxies), interest rate (various proxies), stock market
Chen et al. (1998)
Regression Inflation (various proxies), interest rate (various proxies), stock market
Chen et al. (2012)
Regression Interest rate
Downs et al. (2003) VAR (Variance decomposition and impulse response)
Stock market, interest rate (various proxies), other variables (construction, industrial production, dividend yield)
Ewing & Payne (2005) VAR (generalised impulse response)
Interest rate (federal funds rate, the default risk premium) the index of coincident indicators, inflation
Glascock, Lu & So (2002) VECM, VAR (variance decomposition and impulse response)
Interest rate (federal fund rate), inflation (CPI, unexpected, expected), other variables (industrial production)
Huang & Lee (2009) Autoregressive Integrated Moving Oil, stock market, interest rate (short
44
Average (ARIMA) term and long term).
He, Webb & Myer (2003)
Flexible least square Interest rate (various)
Hoesli, Lizieri & MacGregor (2008)
Error correction model Inflation (various)
Laopodis (2009) VAR (variance decomposition and Impulse Response)
Stock market, other variables (industrial production)
Lieser & Groh (2013) Augmented panel regression analysis
Other variables (economic activities, real estate investment opportunities, depth and sophistication of capital markets, investor protection and legal framework, admin burden and regulatory limitation, socio-cultural and political environment)
Ling & Naranjo (1997) Regression Inflation, interest rate (treasury bill, term structure), other variables (consumption expenditures)
Liow (2010) Cointegration test Stock market
Mueller & Pauley (1995) Regression Stock market, interest rate (various)
Nittayagasetwat & Buranasiri (2012)
Regression Interest rate (various)
Okunev, Wilson & Zurbruegg (2000)
VAR granger (causality) Stock market
Payne (2003) VAR (impulse response) Stock market, inflation, interest rate (various), other variables (industrial production), inflation
Simpson, Ramchander & Webb (2007)
Regression Inflation (various)
Swanson, Theis & Casey (2002)
Regression Interest rate (various)
Yunus (2012) Cointegration and VAR (Granger causality and impulse response)
Stock, economic growth, inflation, interest rate, other variable (money supply)
45
3.1.1 Regression analysis
Regression analysis is the main tool of econometrics and concerned with the statistical
dependence of the dependent variable on one or more explanatory variables (Gujarati 2003).
This method and its modified approaches are a popular method known as cross-sectional
regression, as the explanatory variables are associated with one period or point in time. The
regression is subject to normal distribution of the data, and thus the Jarque–Bera test was
applied to test stationarity (Clark & Daniel 2006).
This method were used in a number of prior studies (see Chen, Roll & Ross 1986; Mueller &
Pauley 1995; Chen, Hsieh & Jordan 1997; Ling & Naranjo 1997; Chatrath & Liang 1998; Chen et
al. 1998; Clayton & MacKinnon 2001; Swanson, Theis & Casey 2002; He, Webb & Myer 2003;
Standish et al. 2005; Clark & Daniel 2006; Simpson, Ramchander & Webb 2007; Franken, Bloom
& Erasmus 2011; Adrangi, Chatrath & Raffliee 2004; Chen et al. 2012; Nittayagasetwat &
Buranasiri 2012; Simo-Kengne, Bittencourt & Gupta 2012).
The modified regression approaches, such as flexible least square (He, Webb & Myer 2003),
seemingly unrelated regression (Simo-Kengne, Bittencourt & Gupta 2013), augmented panel
regression analysis (Lieser & Groh 2013), were not frequently used and are not applicable to
this research; thus, they were not applied.
The assumptions of the multiple regression models are (Pindyck & Rubinfeld 1998):
1. The relationship between the independent (X) and the dependent (Y) variables is linear.
2. The independent variable (X) has no exact linear relationships between two or more
independent variables, i.e., the independent variables are independent.
3. The error has zero expected value for all observations.
4. The error terms have constant variance for all observations. If not, it is called
heteroskedasticity.
5. Errors terms corresponding to different observations are independent and therefore
uncorrelated.
6. The error term is normally distributed.
46
The formula for regression will be as follows:
Equation 3: Regression
Y = α + β1 X1 + β2 X2+ … βn Xn+ ε
[3.1]
Where:
Y is the dependent variable, α is a constant, β1, 2, n are coefficient of independent variables, X1,
2, n are independent variables and ε is error term.
More specifically, the regression equation would be:
Equation 4: Regression of Unlisted Property
IPD = α + β1 ALSI + β2 CPI + β3 GDP + β4 LB + ε
[3.2]
Where:
IPD IPD total return (annual)
ALSI Changes in JSE/FTSE All-Share Index total return (i.e., include dividends, annual)
CPI Changes in Consumer Price Index (annual)
GDP Changes in Gross Domestic Product (annual)
LB Changes in 10-Year Government Yield (annual)
47
Equation 5: Regression of Listed Property
J256T = α + β1 ALSI + β2 CPI + β3 GDP + β4 LB + ε
[3.3]
Where:
J256T The changes in total returns indices of Property Loan Stock (include all
distribution, quarterly)
ALSI Changes in JSE/FTSE All-Share Index total return (i.e., include dividends,
quarterly)
CPI Changes in Consumer Price Index (quarterly)
GDP Changes in Gross Domestic Product (quarterly)
LB Changes in 10-Year Government Yield (quarterly)
3.1.2 Vector autoregression (VAR)
Vector autoregression was introduced as an alternative approach to multi-equation modelling
through the work of Christopher Sims (Pindyck & Rubinfeld 1998), who was awarded Nobel
Memorial Prize in Economics in 2011 for his ‘empirical research on cause and effect in the
macroeconomy’.15
Sims (1980) formulated the VAR model to assume all variables to be endogenous. The
application of VAR model requires only two specifications:
1. The variables (endogenous and exogenous) are believed to interact and hence are
included as part of the economic system.
15 Nobel Prize Website – http://www.nobelprize.org/nobel_prizes/economic-sciences/laureates/2011/sims-facts.html
48
2. The largest number of lags needed must capture most of the effects that the variables
have on each other (Pindyck & Rubinfeld 1998).
The VAR methodology has the following advantages over other methods:
1. It has few theoretical restrictions (Laopodis 2009).
2. It enables the researcher to determine the adjustment time required for REIT to
incorporate information from change within these economic variables (Downs et al.
2003).
3. It provides information on the magnitude of the shocks.
4. It is suitable when variables within the model are highly autocorrelated.
Numerous prior studies in this field have applied this methodology and its derivatives (see
Brooks & Tsolacos 1999; Glascock, Lu & So 2002; Downs et al. 2003; Payne 2003; Ewing & Payne
2005; Laopodis 2009; Chang, Chen & Leung 2011; Boudry, Coulson, Kallberg & Liu 2012; Yunus
2012; Simo-Kengne, Bittencourt & Gupta 2013).
There are three forms of VAR model, namely: reduced, structured, and recursive (Stock &
Watson 2001).
A reduced form of VAR expresses each variable as a linear function of its own past
values, the past values of all other variables being considered and a serially uncorrelated
error term.
A structural VAR uses economic theory to sort out the contemporaneous links among
the variables. Structural VARs require ‘identifying assumptions’ that allow correlations
to be interpreted causally. These identifying assumptions can involve the entire VAR, so
that all of the causal links in the model are spelled out, or just a single equation, so that
only a specific causal link is identified.
A recursive VAR constructs the error terms in each regression equation to be
uncorrelated with the error in the preceding equations. This is done by judiciously
including some contemporaneous values as regressors.
49
This research applies the reduced form, as the study assumes no structural inferences of the
variables (i.e., does not impose any restrictions about which of the variables affect the others,
as would be the case in a regression model). It is assumed that variables are related to their
own lagged values and the lagged values of the other variables over time (Brooks & Tsolacos
1999).
The general form of a VAR model is given by the following unrestricted reduced-form) system:
Equation 6: General form of VAR
Yt = α + β (L) Zt + vt [3.4]
Where:
Yt is a vector of the n stationary endogenous variable, α is an n X 1 vector of constants, β (L) is
an n X n matrix of (lagged) polynomial coefficients, and vt is an n X 1 vector of white noise
innovation terms with E (vtk) = 0 and E (vtk , vsk) = 0 for t ≠ s. The disturbance term vt, also has a
covariance matrix, E ( vt, vt)= ∑.
Finally, the lag operator is defined as: β (L) = β 1 + β 2 + … + β k Lk-1 of degree k -1 and β j,
for j =1, …k. Laopodis (2009)
Specifically, the general four-equation VAR system can be expressed as follows:
Equation 7: VAR of Unlisted Property
IPDt = α +∑ 𝑎1, 𝑖𝑛𝑖=1 IPDt-i + ∑ 𝑏1, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐1, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑1, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒1, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.4.1]
ALSIt = α +∑ 𝑎2, 𝑖𝑛𝑖=1 IPDt-i + ∑ 𝑏2, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐2, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑2, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒2, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.4.2]
CPIt = α +∑ 𝑎3, 𝑖𝑛𝑖=1 IPDt-i + ∑ 𝑏3, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐3, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑3, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒3, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.4.3]
50
GDPt = α +∑ 𝑎4, 𝑖𝑛𝑖=1 IPDt-i + ∑ 𝑏4, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐4, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑4, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒4, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.4.4]
LBt = α +∑ 𝑎5, 𝑖𝑛𝑖=1 IPDt-i + ∑ 𝑏5, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐5, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑5, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒5, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.4.5]
Where:
IPD IPD total return (annual)
ALSI Changes in JSE/FTSE All-Share Index total return (i.e., include dividends, annual)
CPI Changes in Consumer Price Index (annual)
GDP Changes in Gross Domestic Product (annual)
LB Changes in 10-Year Government Yield (annual)
Equation 8: VAR of Listed Property
J256Tt = α +∑ 𝑎1, 𝑖𝑛𝑖=1 J256Tt-i + ∑ 𝑏1, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐1, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑1, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒1, 𝑖𝑛𝑖=1 LBt-i
+ ε [Formula 3.5.1]
ALSIt = α +∑ 𝑎2, 𝑖𝑛𝑖=1 J256Tt-i + ∑ 𝑏2, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐2, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑2, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒2, 𝑖𝑛𝑖=1 LBt-i +
ε [Formula 3.5.2]
CPIt = α +∑ 𝑎3, 𝑖𝑛𝑖=1 J256Tt-i + ∑ 𝑏3, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐3, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑3, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒3, 𝑖𝑛𝑖=1 LBt-i +
ε [Formula 3.5.3]
GDPt = α +∑ 𝑎4, 𝑖𝑛𝑖=1 J256Tt-i + ∑ 𝑏4, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐4, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑4, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒4, 𝑖𝑛𝑖=1 LBt-i +
ε [Formula 3.5.4]
LBt = α +∑ 𝑎5, 𝑖𝑛𝑖=1 J256Tt-i + ∑ 𝑏5, 𝑖𝑛
𝑖=1 ALSIt-i + ∑ 𝑐5, 𝑖𝑛𝑖=1 CPIt-i + ∑ 𝑑5, 𝑖𝑛
𝑖=1 GDPt-i + ∑ 𝑒5, 𝑖𝑛𝑖=1 LBt-i + ε
[Formula 3.5.5]
51
Where:
J256T Changes in total returns indices of Property Loan Stock (include all distribution,
quarterly)
ALSI Changes in JSE/FTSE All-Share Index total return (i.e., include dividends,
quarterly)
CPI Changes in Consumer Price Index (quarterly)
GDP Changes in Gross Domestic Product (quarterly)
LB Changes in 10-year Government Yield (quarterly)
The VAR tests included in this study are vector autoregression, variance decomposition and
impulse response functions.
The variance decomposition expresses each variable mathematically as a linear combination of
its and other variables’ current and past forecast errors (residual terms) (Downs et al. 2003).
The impulse response function of the VAR analysis provides insight on the speed of information
transmission among the commercial property returns and the economic variables. Also its
analysis identifies changes over time in the dependent variables.
All variables included in the VAR tests need to be stationary in order to carry out the
significance test. Thus, all variables were subjected to augmented Dickey–Fuller tests. Also, in
order for a VAR to be unrestricted, it is required that the same number of lags of all of the
variables is used in all equations. Therefore, Akaike’s information criterion (AIC) was used.
52
3.2 Data and proxy selection
Macroeconomic variables relevant to this study were selected from prior research (see Chapter
2). In addition to identifying potential macroeconomic variables that influence property returns,
relevant proxies for each of the variables were identified.
3.2.1 Unlisted properties
The commercial properties returns data used as the proxy of unlisted commercial properties
were collected from Investment Property Databank (IPD). IPD published the first index in 1997.
As of the end of 2011, the IPD databank in South Africa includes over 2000 properties with a
value of over R205 billion16. IPD returns show the return of direct property without any gearing
or market impacts. The observation of the data is on an annual basis, as the IPD data is
available annually.
Earlier studies by Hoesli, Lizieri and MacGregor (2008), Boudry et al. (2012) and Towers Watson
(2012) used IPD data.
The other proxies identified for analysis were: stock market, economic growth, inflation and
interest rates.
Stock market: The proxy chosen is the annual percentage change FTSE/JSE All-Share Index
(J203T), which is the total returns index (includes dividends distribution). The data was obtained
from INET BFA17. The use of J203T is a consistent with prior local research such as Clark and
Daniel (2006) and Franken, Bloom and Erasmus (2011). However, this research uses total
returns are used instead of price only, as most prior research locally and internationally uses
price indices only.
Economic growth: The proxy chosen was the annual percentage change in GDP. The data was
obtained from Bureau of Economic Research (BER). GDP was used as a proxy in several studies
16 www.ipd.co.za 17 www.inetbfa.com
53
(see Standish et al. 2005; Clark & Daniel 2006; Chang, Cheng & Leung 2011; Franken, Bloom &
Erasmus 2011; Boshoff & Cloete 2012; Yunus 2012; Lieser & Groh 2013).
Inflation: The proxy chosen is the annual percentage change of the annual CPI. The data was
calculated by Statistics South Africa and obtained from BER. CPI is a consistent proxy for
inflation as per previous international studies (see Chen, Roll & Ross 1986; Yobaccio, Rubens &
Ketcham 1995; Chen, Hsieh & Jordan 1997; Chatrath & Liang 1998; Chen et al. 1998; Glascock,
Lu & So 2002; Payne 2003; Adrangi, Chatrath & Raffliee 2004; Ewing & Payne 2005; Yunus
2012). In South African literature, Firer and McLeod (1999), Hassan and Biljon (2009) and
Franken, Bloom and Erasmus (2011) use CPI.
Interest rate: The proxy chosen is the annual movement of ten-year government bond yield.
The interest rate proxy is nominal, i.e., not adjusted with inflation (see Brooks & Tsolacos 1999).
The data was maintained by South African Reserve Bank and obtained from BER.
The ten-year government bond yield is chosen over the Prime Interest Rate for the following
reasons:
1. Prime Interest Rate is a borrowing interest rate for investors (i.e., investors cannot
invest in the Prime Interest Rate). Thus in terms of a theoretical framework, it is not
relevant for the estimation of cost of capital and capitalisation rate.
2. The government bond yield is considered as an investable risk-free return. As listed
properties as an asset class are often compared to government bond, this can be
observed from the publications from numerous investment analysts and asset
managers (see Appendix 1).
3. Prior research in South Africa used the Prime Interest Rate as an interest rate proxy, as
the studies all dealt with residential properties, and borrowing costs in residential
properties are generally linked; thus Prime Interest Rate is a relevant proxy for
residential (see Clark & Daniel 2006; Franken, Bloom & Erasmus 2011)
4. According to PwC (2012), the proxies used by business valuation practitioners for risk
free rate are predominantly South African government bonds (79% of respondents in
54
2012, and in the 2010 survey, 93% of the respondent used South African government
bonds).
Table 12 summarises the variables and proxies used for the unlisted properties.
Table 12: The variables and proxies used
Variables Proxy (Code) Data period Frequency Source
Unlisted commercial
property returns
Total return of IPD databank
(IPD)
1995 – 2012 Yearly IPD
Stock market Percentage changes of All-
Share Index Total Returns
(ALSI)
1995 – 2012 Yearly INET BFA
Economic growth Percentage changes in GDP (%
GDP)
1995 – 2012 Yearly BER
Inflation Percentage changes in
Consumer Price Index (CPI)
1995 – 2012 Yearly BER
Interest rates Percentage changes in 10-
Year Government Yield (LB)
1995 – 2012 Yearly BER
3.2.2 Listed properties
For securitised commercial properties, the proxy used is the quarterly changes of total return
indices of PLS (J256T). PUT was not included in this study, as it accounted for only a small
percentage of the listed property sector (see Figure 1 in Chapter 1, PUT is in blue and PLS in
red).
The total return index is used, as this index captures the income return as well as the capital
return of listed property. Income is an important part of securitised commercial properties
return and thus must be incorporated. Figure 3 shows the historical total quarterly returns for
PLS Income return (in red) and price appreciation (in blue). Several studies used total return
indices: Brooks and Tsolacos (1999); Glascock, Lu and So (2002); Okunev, Wilson and Zurbruegg
(2000); Laopodis (2009); and Nittayagasetwat and Buranasiri (2012). The other prior studies
cited did not specify whether total return indices were used.
55
The data is quarterly, as the GDP data is available quarterly.
Figure 3: Quarterly Return of PLS – August 2013 (SA Reit Association 2013a)
The following proxies were identified for analyses:
Quarterly percentage change FTSE/JSE All-Share Index (for stock market)
Quarterly percentage change in the Gross Domestic Product (for economic growth)
Quarterly percentage change in Consumer Price Index (for inflation)
Quarterly percentages movement of averages of monthly ten-year government bond
yield (for interest rates)
The proxies were the same as the proxies used for IPD, except the period is from 2002 to 2013,
and the frequency is quarterly. This is due to the fact that J256T was started in 2002, as in that
year the FTSE Group and the JSE entered into a partnership to create the FTSE/JSE Africa index
Series. These new indices, which apply the FTSE global classification system (Mangani 2011),
replaced the old indices and brought about a change in calculation methodologies18; limited
18 See http://www.jse.co.za/Libraries/JSE_Magazine_educational_pullouts/Indices_for_Beginners.sflb.ashx
56
data were available prior to 2002. Table 13 summarises the variables and proxies used for the
listed properties.
Table 13: Summary of variables, proxies, period and source of data
Variables Proxy Data period Frequency Source
Listed Commercial
Property Returns
Total return of FTSE/JSE
Listed Property Loan Stock
(J256T)
Sept 2002 to
Sept 2013
Quarterly INET BFA
Stock Market All-Share Index total
returns (ALSI)
Sept 2002 to
Sept 2013
Quarterly INET BFA
Economic Growth Percentage changes in GDP
(% GDP)
Sept 2002 to
Sept 2013
Quarterly BER
Inflation Percentage changes in
Consumer Price Index (CPI)
Sept 2002 to
Sept 2013
Quarterly BER
Interest rates Percentage changes in 10-
Year Government Yield
(LB)
Sept 2002 to
Sept 2013
Quarterly BER
3.2.3 Comparison between the proxy of listed and unlisted properties
According to Stan Garrun of IPD, IPD covers 65% of professionally managed unlisted properties
and 80% of the listed property sector. The effect of the overlap is evident from Figure 4, which
graphs the movement of ALSI, IPD and J256T from 2002 to 2012. Table 14 demonstrates
correlation between the three variables ranges from 62% to 77%. The IPD is 69% correlated to
J256T.
Table 14: Summary of variables, proxies, period and source of data
J256T ALSI IPD
J256T 100%
ALSI 77% 100%
IPD 69% 62% 100%
The differences can be attributed to three variables:
57
1. Difference in population: not all listed property funds report to IPD, and IPD covers
some unlisted property fund as well.
2. Gearing effects: IPD reports asset returns of underlying property; however, PLS
companies in J256 are allowed to borrow and thus may be able to generate higher
returns through gearing.
3. Difference in capital returns: IPD is an appraisal-based index and is not transaction-
based. However, capital returns are based on the appreciation of the linked unit and do
not necessarily correlate to the performance of the underlying assets. (For instance
Growthpoint’s trading price may exceed the Net Asset Value of the stock, due to the
nature of the market.)
Figure 4: Returns of PLS, IPD and ALSI
This result is consistent with prior research by Boshoff and Cloete (2012) and Boudry et al.
(2012), which found listed property relating to the underlying property market and activities.
-30%
-20%
-10%
0%
10%
20%
30%
40%
50%
60%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Re
turn
s p
.a.
J256T
ALSIT
IPD
58
3.2.4 Other proxies considerations
The variables used in this research are basic series and not derived series. Some examples of
derived series used in prior research include:
Real returns, i.e., nominal returns less inflation (Adrangi, Chatrath & Raffliee 2004).
Excess return (Chen, Roll & Ross 1986; Ling & Naranjo 1997; Payne 2003; Laopodis 2009;
Chen et al. 2012; Nittayagasetwat & Buranasiri 2012).
Expected inflation (Chen, Roll & Ross 1986; Chen, Hsieh & Jordan 1997; Ling & Naranjo
1997; Chen et al. 1998; Simpson, Ramchander & Webb 2007; Hoesli, Lizieri &
MacGregor 2008)
Unexpected inflation, defined as the difference between the realised inflation rate (as
measured by the Consumer Price Index) and the expected inflation rate (Chen, Roll &
Ross 1986; Ling & Naranjo 1997; Brooks & Tsolacos 1999; Simpson, Ramchander &
Webb 2007; Hoesli, Lizieri & MacGregor 2008)
Real interest, defined as nominal treasury bill rate less inflation (Chen, Roll & Ross 1986;
Standish et al. 2005)
Risk premium, defined as Baa bond yield less long-term government yield (Chen, Roll &
Ross 1986; Payne 2003; Chen et al. 2012; Nittayagasetwat & Buranasiri 2012)
Term structure, defined as difference between long-term government yield less the
Treasury bill rate (Chen, Roll & Ross 1986; Ling & Naranjo 1997; Chen, Hsieh & Jordan
1997; Chen et al. 1998; Swanson, Theis & Casey 2002; He, Webb & Myer 2003; Payne
2003; Nittayagasetwat & Buranasiri 2012)
Property returns residual (employed by Brooks & Tsolacos 1999); the property returns
were regressed on the stock market index and residuals saved.
This research avoided using the derived series, as the purpose of this research is to test the
direct relationships between returns and identified economic variables. Clark and Daniel (2006),
Laopodis (2009), Franken, Bloom and Erasmus (2011), and Yunus (2012) applied the
methodology of using the basic series.
59
Industrial production is another indicator of economic growth (used by Chen, Roll & Ross 1986;
Adrangi, Chatrath & Raffliee 2004; Ling & Naranjo 1997; Glascock, Lu & So 2002; Downs et al.
2003; Payne 2003; Laopodis 2009; Chen et al. 2012). Ewing and Payne (2005) used a
coincidental index.
3.3 Application of statistical methodology
Eview19 software was used to perform the statistical analysis. The process of statistical analysis
was as follows:
1. Collect raw data described in section 3.2
2. Process data – as some data collected are in index form and thus need to be converted
into ‘changes in’ the relevant proxies20.
3. Data analysis
a. Sample period analysis – to consider the business cycle and interest rate cycle of
the sample period
b. Descriptive statistics
i. Histogram – use to detect outliers
ii. Descriptive statistics – brief summary of the data
iii. Jarque–Bera test - ensure the data is normally distributed
iv. Correlation matrix – to identify the relationships between the variables
c. Regression analysis
i. Results
ii. Assumptions – ensuring the assumptions for multiple regression are met
d. Vector autoregression
i. Histogram – used to detect outliers
19 Version 7, a statistical package for Windows, used mainly for time-series oriented econometric analysis, http://www.eviews.com/ 20 During this process, the author detected that some of the data provided by INET BFA was incorrect. INET BFA was informed and data was subsequently corrected.
60
ii. Dickey–Fuller test – ensure the statistical process is stationary
iii. Vector autoregression
iv. Variance decomposition – The variance decomposition expresses each
variable mathematically as a linear combination of its and other
variables’ current and past forecast errors.
v. Impulse response – The impulse response function of the VAR analysis
provides insight on the speed of information transmission among the
commercial property returns and the economic variables.
3.4 Summary
This chapter presented the two main statistical methodologies:
Regression analysis and its underlying assumptions
Vector autoregression and its underlying assumptions
Further, the chapter discussed the detail, source and justification for each of the proxies used.
Moreover, it considered the data that would be appropriate in a South African context,
including proxies for that could be applied for both listed and unlisted properties. The chapter
tested a comparison between the performance of listed and unlisted property for the period
2002 to 2012. It found that the performance of listed and unlisted property was highly
correlated. In addition, this chapter listed proxies used in previous studies that were excluded
from this study. Finally, this section concluded the chapter with a description of each of the
steps of the statistical methodology to be applied in the following chapter.
61
4. Results and Discussion
This chapter presents the results of the statistical analysis performed, the interpretations and
the discussion of the results.
4.1 Unlisted properties (IPD)
4.1.1 Sample period analysis
Figure 5 shows the sample size of the data set is from 1995 to 2012 (18 annual data points per
series). This period is characterised by a declining interest rate from 16% to 8%, thus not one
full interest rate cycle. However, over the same period, there is at least one and a half business
cycles (Composite Business Cycle – leading indicator21 is used as a proxy for business cycles).
Business and interest rate cycles were considered, as the period of research may affect the
outcome of the results. For instance, Chen et al. (2012) found that monetary policy (the fed rate
is used as a proxy) has a differential impact on the REIT market, and that the impact is
significantly negative during bull markets and has no impact during bear/volatile markets.
21 compiled by the South African Reserve Bank since 1983
62
Figure 5: Business cycle and ten-year bond rate over the sample period
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
-20%
-15%
-10%
-5%
0%
5%
10%
15%
20%
25%
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Composite Business Cycle Indicators - Leading Indicator (Year on Year %change)
10 Year Bond Rate
Interest Rate %Business Cycle YOY %
63
4.1.2 Descriptive statistics
Histogram
Histograms are used in Figure 6 to detect outliers (defined as extreme values that are very far
removed from the rest of the data set) in the data set (Gujarati 2003). After reviewing the
histograms, no outliers were detected at the either end of the histogram and none were
removed from the data set.
Figure 6: Data of variables
0
1
2
3
4
5
-.4 -.2 .0 .2 .4 .6 .8
Freq
uenc
y
ALSIT
0
1
2
3
4
5
-.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 .05
Freq
uenc
y
CPI
0
1
2
3
4
5
-.06 -.05 -.04 -.03 -.02 -.01 .00 .01 .02 .03 .04 .05
Freq
uenc
y
GDP
0
1
2
3
4
5
.00 .04 .08 .12 .16 .20 .24 .28 .32
Freq
uenc
y
IPD
0
1
2
3
4
5
6
7
-.025 -.020 -.015 -.010 -.005 .000 .005 .010 .015
Freq
uenc
y
LB
64
Descriptive statistics
Table 15 summarises the descriptive statistics used.
Table 15: Descriptive statistics
IPD ALSI CPI GDP LB
Mean 0.16 0.17 0.00 0.00 0.00
Median 0.14 0.18 0.00 0.00 0.00
Maximum 0.30 0.71 0.04 0.05 0.01
Minimum 0.05 -0.23 -0.04 -0.05 -0.02
Std. Dev. 0.07 0.23 0.03 0.02 0.01
Skewness 0.82 0.44 -0.11 -0.21 -0.32
Kurtosis 2.66 2.97 2.21 3.86 3.01
Observations 18 18 18 18 18
Mean: Also known as average (Gujarati 2003), the mean is the weighted average of the data.
The annual average total return for IPD during this period is 16% and ALSI is 17%.
Median: This is the number located in the middle of the data set (UCT 2011). The annual
median total return for IPD during this period is 14% and ALSI is 18%; they are very close to the
mean.
Maximum: This is the highest value in the data set; the highest annual return for IPD was 30%
and ALSI is 71%22.
Minimum: This is the lowest value in the data set. The lowest annual return for IPD was 5% and
for ALSI is -23%.
Standard deviation: This is the square root of variance and is the most commonly employed
measure of spread (UCT 2011). Higher numbers indicate a higher spread from the mean.
Skewness: This is a statistic that provides useful information about a symmetry probability
distribution. For all symmetric distribution skewness should be zero. For non-symmetric
distribution, it is positive when upper tail is thicker than lower tail and vice versa (Pindyck &
Rubinfeld 1998). In the results, IPD and ALSI are marginally positive, and CPI, GDP, and LB are
marginally negative.
22 This number was verified against other data sources and is correct.
65
Kurtosis: This is a measure of ‘thickness’ of the tail distribution; for a normal distribution it is
three. If the results are greater than three, it is thicker and vice versa (Pindyck & Rubinfeld
1998). IPD, ALSI and CPI are just below three; GDP and LB are above three.
Observation: This is the sample size of the data set. In this case, it is annually from 1995 to 2012
(18 data points per series).
Jarque–Bera test
Regression is a parametric test; thus, the data must be tested using the Jarque–Bera test to
ensure the data is normally distributed. Jarque-Bera measures the difference of the skewness
and kurtosis of the series with those from the normal distribution. Under the null hypothesis of
a normal distribution, the Jarque–Bera statistic is distributed X2 as with two degrees of freedom
(Pindyck & Rubinfeld 1998).
Table 16: Jarque–Bera test
IPD ALSI CPI GDP LB
Jarque–Bera 2.13 0.59 0.50 0.68 0.30
Probability 0.35 0.74 0.78 0.71 0.86
Reject Null hypothesis at 5% No No No No No
Based on the results of Jarque–Bera test in Table 16, the data series can be assumed to be
normally distributed.
66
Correlation matrix
Table 17: Correlation matrix
IPD ALSI CPI GDP LB
IPD 1.00 0.42 0.20 0.24 -0.08
ALSI 0.42 1.00 -0.34 0.17 -0.36
CPI 0.20 -0.34 1.00 -0.04 0.18
GDP 0.24 0.17 -0.04 1.00 0.03
LB -0.08 -0.36 0.18 0.03 1.00
In the correlation matrix (Table 17) above, IPD returns are positively correlated to ALSI, CPI and
GDP, and are negatively related to the ten-year bond rate. None of the independent variables
are correlated for greater than 0.7 or less than -0.7; thus there is no concern with
multicollinearity (a statistical phenomenon in which two or more independent variables in a
multiple regression model are highly correlated).
4.1.3 Regression analysis and vector autoregression
Due to the limited sample size of IPD data (18 data points), the results of the statistical analysis
such as regression and the VAR model do not have any statistical significance.
There is no general description on what the minimum sample size should be, as it is dependent
on sample distribution and a number of independent variables. Some economists think 30 is
sufficient (Wooldridge 2012, p. 176)
4.2 Listed (J256T, Property Loan Stocks)
4.2.1 Sample period analysis
Figure 7 shows the data set is quarterly from 2002 to 2013 (46 data points per series). This
period is essentially not one full interest-rate cycle but half a cycle of declining interest rates
67
(from 12% to 8%). However, over the same period, there is at least one business cycle (the
Composite Business Cycle – leading indicator23 is used as a proxy for business cycles).
Figure 7: Business cycle and ten-year bond rate over the sample period
23 Compiled by South African Reserve Bank since 1983
0%
2%
4%
6%
8%
10%
12%
14%
-20%
-15%
-10%
-5%
0%
5%
10%
15%
20%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Composite Business Cycle Indicators - Leading Indicator (Year on Year %change)10 Year Bond Rate
Interest Rate %Business Cycle YOY %
68
4.2.2 Descriptive statistics
Histogram
As shown in Figure 8, no outliers (defined as extreme values that are very far removed from the
rest of the data set) were detected from the histograms earlier, and none were removed from
the data set (Gujarati 2003). All the variables appear to be stationary.
Figure 8: Data of Variables
0
2
4
6
8
-.25 -.20 -.15 -.10 -.05 .00 .05 .10 .15 .20 .25
Freq
uenc
y
ALSIT
0
2
4
6
8
10
12
-.04 -.03 -.02 -.01 .00 .01 .02 .03 .04
Freq
uenc
y
CPI
0
1
2
3
4
5
6
7
-.08 -.06 -.04 -.02 .00 .02 .04 .06
Freq
uenc
y
GDP
0
2
4
6
8
-.20 -.15 -.10 -.05 .00 .05 .10 .15 .20 .25
Freq
uenc
y
J256T
0
2
4
6
8
10
12
-.016 -.012 -.008 -.004 .000 .004 .008 .012 .016
Freq
uenc
y
LB
69
Descriptive statistics
Table 18 summarises the descriptive statistics data set.
Table 18: Descriptive statistics
J256T ALSI CPI GDP LB
Mean 0.06 0.04 <0.001 0.01 <0.001
Median 0.06 0.06 0.001 0.02 0.00
Maximum 0.25 0.20 0.03 0.04 0.01
Minimum -0.20 -0.21 -0.04 -0.06 -0.01
Std. Dev. 0.09 0.08 0.01 0.03 0.01
Skewness -0.46 -0.82 -0.71 -0.92 0.30
Kurtosis 3.59 3.72 3.57 2.81 2.55
Observations 46 46 46 46 46
Mean: Also known as the average (Gujarati 2003), the mean is the weighted average of the
data. The quarterly average return for J256T during this period is 6% and ALSI is 4%.
Median: This is the number located in the middle of the data set (UCT 2011). The annual
median total return for J256T during this period is 6% and ALSI is 6%; thus they are very close to
the mean.
Maximum: This is the highest value in the data set. The highest quarterly return for J256T was
25% and ALSI was 20%.
Minimum: This is the lowest value in the data set. The lowest annual return for J256T was -20%
and for ALSI is -21%.
Standard deviation: This is the square root of variance and is the most commonly employed
measure of spread (UCT 2011). Higher numbers indicate a higher spread from the mean.
Skewness: This is a statistic that provides useful information about symmetry of probability
distribution. For all symmetric distribution skewness should be zero. For non-symmetric
distribution, it is positive when the upper tail is thicker than the lower tail and vice versa
(Pindyck & Rubinfeld 1998). In this case, J256T, ALSI, CPI and GDP’s lower tails are slightly
thicker than their upper tails.
70
Kurtosis: This is a measure of ‘thickness’ of the tail distribution; for normal distribution it is
three. If the results are greater than three, it is thicker and vice versa (Pindyck & Rubinfeld
1998). IPD, ALSI and CPI are all just above three; GDP and LB are just below three.
Observation: This is the sample size of the data set. In this case it is quarterly from 2002 to 2013
(46 data point per series).
Jarque–Bera test
Based on the results of Jarque–Bera test in Table 19 the data series can be assumed to be
normally distributed.
Table 19: Jarque–Bera test
J256T ALSI CPI GDP LB Jarque–Bera 2.29 6.13 4.45 6.45 1.07
Probability 0.32 0.05 0.11 0.04 0.59
Reject Null hypothesis at 5% No No No No No
Correlation matrix
Table 20: Correlation matrix
J256T ALSI CPI GDP LB J256T 1.00 0.14 -0.07 -0.07 -0.56
ALSI 0.14 1.00 -0.14 0.13 0.04
CPI -0.07 -0.14 1.00 0.17 0.22
GDP -0.07 0.13 0.17 1.00 0.34
LB -0.56 0.04 0.22 0.34 1.00
The correlation matrix above (Table 20) demonstrated that J256T returns are positively
correlated to ALSI and negatively related to GDP, CPI and ten-year bond rate. None of the
independent variables are correlated for greater than 0.7 or less than -0.7; thus there is no
71
concern with multicollinearity (a statistical phenomenon in which two or more independent
variables in a multiple regression model are highly correlated). The correlation results for ALSI
and ten-year bond are consistent with unlisted property returns.
4.2.3 Regression analysis
Regressions results (all variables): Initially without any adjustments, the Durbin–Watson (DW)
test for the regression had a result of 2.68, indicating a negative serial correlation. This means
that errors corresponding to different observations are not independent and are therefore
correlated (Pindyck & Rubinfeld 1998), and thus violated an assumption required for multiple
regression. Thus, autoregressive function, ar (1) was subsequently incorporated into the
regression to adjust for the negative serial correlation, after the incorporation of ar (1); the DW
stat is around two (see Table 21).
Table 21: Regression results after incorporation of autoregressive function (all
variables)
Variable Coefficient Std. Error t-Statistic Prob.
C 0.04 0.01 3.80 <0.001
ALSI 0.21 0.12 1.81 0.08
CPI 0.55 0.67 0.82 0.42
GDP 0.78 0.51 1.53 0.13
LB -13.19 2.21 -5.97 <0.001
AR (1) -0.42 0.16 -2.71 0.01
R-squared 0.46
Adjusted R-squared 0.39
S.E. of regression 0.07
Sum squared resid 0.21
Log likelihood 55.10
F-statistic 6.55
Prob (F-statistic) <0.001
Mean dependent var 0.07
S.D. dependent var 0.10
Akaike info criterion -2.23
72
Schwarz criterion -1.99
Hannan-Quinn criter. -2.14
Durbin-Watson stat 2.08
Inverted AR Roots 0.07
This regression model (after incorporating an autoregressive function) can explain 46% (see R-
squared in Table 21) of the J256T’s returns. The ten-year bond was significant at 0% level (see
probability) and ALSI was significant at 8% level (see Table 21). CPI and GDP were not found to
be statistically significant. The coefficient of ALSI was positive, indicating a positive relationship
between ALSI and J256T. The negative coefficient of LB indicated a negative relationship
between LB and J256T. The results are consistent with correlation matrix in section 4.2.2.
The F-statistical probability indicates this regression model is statistically significant at less than
0.1%.
Stepwise regression
Researchers often use stepwise regression to determine the best explanatory variables for a
regression model (Gujarati 2003). By using stepwise regression, the ten-year government bond
(probability at less than 0.1%) and the All-Share Index (probability at 2%) were identified as
significant variables at 5% level (see Table 22). The coefficient was negative for LB and positive
for ALSI, consistent with previous regressions.
Table 22: Stepwise regression results
Variable Coefficient Std. Error t-Statistic Prob.*
LB -13.01 2.63 -4.94 <0.001
ALSI 0.35 0.14 2.44 0.02
GDP 0.83 0.51 1.65 0.11
R-squared 0.19 Mean dependent var 0.06
Adjusted R-squared 0.15 S.D. dependent var 0.09
S.E. of regression 0.09 Akaike info criterion -1.98
Sum squared resid 0.32 Schwarz criterion -1.86
Log likelihood 47.48 Hannan-Quinn criter. -1.93
Durbin–Watson stat 2.41
73
Regressions results (two variables)
Based on the results of the stepwise regression, only two variables were found to be
statistically significant (with ALSI and LB); thus a two-variable model was created.
Initially without any adjustments, the Durbin–Watson test for the regression indicated a
negative serial correlation (with DW statistics of 2.60). The researcher then incorporated an
autoregressive function, ar (1), into the regression. Table 23 presents the results:
Table 23: Regression results after incorporation of autoregressive function (two variables)
Variable Coefficient Std. Error t-Statistic Prob.
C 0.05 0.01 3.74 0.00
ALSI 0.22 0.14 1.55 0.13
LB -11.67 2.67 -4.37 0.00
AR (1) -0.33 0.17 -1.93 0.06
R-squared 0.41
Adjusted R-squared 0.37
S.E. of regression 0.08
Sum squared resid 0.23
Log likelihood 54.69
F-statistic 9.54
Prob (F-statistic) <0.001
Mean dependent var 0.06
S.D. dependent var 0.09
Akaike info criterion -2.25
Schwarz criterion -2.09
Hannan–Quinn criter. -2.19
Durbin–Watson stat 2.08
74
This regression model with ALSI and LB as explanatory variables (after incorporating an
autoregressive function) can explain 41% of the J256T’s returns, and the model is statistically
significant (F-statistical probability at less than 0.1%). The statistically significant (at 0% level)
negative coefficient of LB indicates a negative relationship between LB and J256T, which is
consistent with the correlation matrix, regression model (all variables) and stepwise regression.
The assumptions for this regression model were considered and accounted for (especially
regarding the error terms and explanatory variables). This was to ensure the validity of the
statistical outcome. Based on the relevant test results, the regression model is valid. Table 24
summarises the relevant tests and results.
Table 24: Assumptions for multiple regression
Assumptions of Multiple Regression Models
Relevant Tests Test Results
The relationship between independent (X) and dependent (Y) is linear.
Examine residual plot Residual plot appears to be linear.
The independent (X) variables have no exact linear relationships between two or more independent variables.
If there are, it is called multicollinearity.
Correlation Matrix and VIF None of the independent variables are correlated for greater than 0.7 or less than -0.7.
The error has zero expected value for all observations.
If the error terms do not have zero mean values, the intercept in the regression equation cannot be estimated. However, in reality, the intercept is not important. In addition, the slope coefficients remain unaffected if this assumption is violated. Thus, this assumption is not of concern
75
here (Gujarati 2003).
The error terms have constant variance for all observations.
Heteroscedasticity is adjusted in the equation by applying the white heteroscedasticity-consistent covariance matrix in the Eview Software.
Errors corresponding to different observations are independent and therefore uncorrelated.
The Durbin–Watson test is the most popular test for testing serial correlation between independent variables.
Autoregressive term is included in the equation.
The error term is normally distributed.
The Jarque–Bera test was tested on the residuals (Jarque–Bera = 0.23, Probability = 0.89).
Cannot reject the hypothesis that the residual is of a normal distribution. Gujarati (2003) contends that the central limit theorem can be relied upon, if the sample size is large enough (30 or more observations). The usual test procedures, namely the t- and F- tests, are still valid even if the error terms are not normally distributed.
4.2.4 Vector autoregression
Histogram
As the data set is the same as in section 4.2.2, we do not have to reperform the analysis.
76
Dickey–Fuller test
The Augmented Dickey–Fuller is used to test the unit root, if there is a unit root and the
statistical process is non-stationary. Thus, a more sophisticated model may need to be adopted.
Table 25 presents the results.
Table 25: Dickey–Fuller test
Variables Probabilities Results
J256T <0.001 Reject hypothesis of unit root
ALSI <0.001 Reject hypothesis of unit root
GDP 0.17 Accept hypothesis of unit root
CPI <0.01 Reject hypothesis of unit root
LB <0.001 Reject hypothesis of unit root
All the series except for GDP did not have a unit root. Thus, the unit root null hypothesis was
rejected, suggesting that the data series can be examined further in this format. The GDP
variables had to be removed from the variables.
Vector autoregression
Two issues are very important when performing VAR analysis: firstly, the ordering of the
variables, and secondly, the appropriate lags must be applied to the model (Laopodis 2009).
There are various ways the variables can be ordered. One method is to rely on economic
theory, and another method is to use statistical techniques. In this study, the results from the
regressions were applied to the ordering of the variables.
In order to determine appropriate lag orders, VAR Lag Order Selection Criteria in Eviews was
used to determine the lag order. Based on the test, the optimal lag is determined to be two.
77
Vector autoregression of the economic variables was performed based on the orders and lag
derived earlier. Table 26 shows the results of the vector autoregression.
Table 26: VAR results
Coefficient Std. Error T-Statistic
LB (-1) 1.12 -4.63 0.24
LB (-2) 0.92 -3.85 0.24
ALSI (-1) 0.16 -0.21 0.76
ALSI (-2) 0.26 -0.21 1.21
CPI (-1) -1.13 -1.62 -0.69
CPI (-2) 1.81 -1.59 1.14
J256T (-1) -0.10 -0.22 -0.44
J256T (-2) -0.17 -0.25 -0.67
C 0.06 -0.03 2.55
R-squared 0.30
Adj. R-squared 0.14
Sum sq. resids 0.00
S.E. equation 0.00
F-statistic 1.87
Log likelihood 177.52
Akaike AIC -7.66
Schwarz SC -7.30
Mean dependent 0.00
S.D. dependent 0.01
78
None of the variables or lags of variables appear to be statistically significant.
The following tests were performed to ensure the fundamental assumptions of VAR met:
1. Testing the unit root of VAR (Figure 9)
2. VAR Residual (Figure 10)
3. VAR Residual Portmanteau Tests for Autocorrelations (Table 27)
The results are as follows:
Figure 9: Testing the unit root of VAR
No root lies outside the unit circle. Thus, VAR satisfies the stability condition.
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Inverse Roots of AR Characteristic Polynomial
79
Figure 10: VAR residual
Based on the charts in Figure 10, it can be observed that the residuals are normally distributed.
The VAR residual portmanteau tests for autocorrelations were perforformed to test whether
there was residual autocorrelation. Based on the results presented in Table 27, there is no
residual autocorrelation.
-.015
-.010
-.005
.000
.005
.010
03 04 05 06 07 08 09 10 11 12 13
LB Residuals
-.20
-.15
-.10
-.05
.00
.05
.10
.15
03 04 05 06 07 08 09 10 11 12 13
ALSIT Residuals
-.03
-.02
-.01
.00
.01
.02
03 04 05 06 07 08 09 10 11 12 13
CPI Residuals
-.3
-.2
-.1
.0
.1
.2
03 04 05 06 07 08 09 10 11 12 13
J256T Residuals
80
Table 27: VAR residual portmanteau tests for autocorrelations
Lags Q-Stat Prob. Adj Q-Stat Prob. df
1 3.08 NA* 3.15 NA* NA*
2 9.88 NA* 10.27 NA* NA*
3 23.61 0.10 25.01 0.07 16.00
4 38.39 0.20 41.26 0.13 32.00
5 49.61 0.41 53.93 0.26 48.00
6 56.09 0.75 61.42 0.57 64.00
7 73.27 0.69 81.86 0.42 80.00
8 82.61 0.83 93.27 0.56 96.00
9 91.34 0.92 104.26 0.69 112.00
10 93.18 0.99 106.63 0.92 128.00
11 108.83 0.99 127.50 0.83 144.00
12 117.34 1.00 139.20 0.88 160.00
*The test is valid only for lags larger than the VAR lag order. df is degrees of freedom for (approximate) chi-square distribution
Based on the tests performed above, the fundamental assumptions of VAR are met and the
results are valid.
81
4.2.4.1 Variance decomposition
Variance decomposition seeks to determine what proportions of the changes in the listed
property return series can be attributed to changes in the lagged explanatory variables (Brooks
& Tsolacos 1999).
Table 28 shows how the economic variables interact with the listed properties index (J256T).
The table presents the percentages of quarterly return variance explained by own volatility and
by the macroeconomic variables.
In the short run (two quarters), the shock to interest rate accounted for 47.4% of the variation
of the fluctuation in the returns of listed properties. Listed property accounted for 46.7% of the
variation of the fluctuation in the returns of listed properties (own shock). The shock to stock
market accounted for 4% and shock to CPI accounted for 1.9% of the variation of the
fluctuation in the returns of listed properties.
In the long run (ten quarters), the shock to interest rate accounted for 46% of the variation of
the fluctuation in the returns of listed properties. Listed property accounted for 43.9% of the
variation of the fluctuation in the returns of listed properties (own shock). The shock to stock
market accounted for 6.1% and shock to CPI accounted for 3.6% of the variation of the
fluctuation in the returns of listed properties.
The results indicated that the volatility of total returns of listed properties is influenced by
interest rates, its own volatility, the stock market and inflation. Interest rates are the greatest
source of volatility. The interest rate variable explained almost 49%, and its own volatility was
almost 48%. The stock market and inflation were not significant contributors to the volatility of
listed properties returns. The significant own shock is similar with the results found by Brooks
and Tsolacos (1999), Downs et al. (2003) and Laopodis (2009); however, in prior studies, own
shock was the largest contributor to variance. In this study, own shock was the second largest
contributor to variance, second to interest rates.
The information content of the economic variables seems to produce effects on the returns
long enough so the quarterly data still expose their influence.
82
Table 28: Variance decomposition
Period S.E. LB ALSI CPI J256T
1 0.00 48.36 3.13 0.81 47.70
2 0.01 47.43 4.00 1.86 46.71
3 0.01 46.43 6.06 3.32 44.19
4 0.01 46.36 6.15 3.35 44.14
5 0.01 46.40 6.11 3.54 43.95
6 0.01 46.39 6.11 3.60 43.91
7 0.01 46.39 6.11 3.60 43.91
8 0.01 46.38 6.11 3.60 43.91
9 0.01 46.38 6.11 3.60 43.91
10 0.01 46.38 6.11 3.60 43.91
Cholesky Ordering: ALSI CPI J256T LB
4.2.4.2 Impulse response
This study conducted an impulse response analysis on the total return volatilities of the listed
properties indices to observe the dynamic interaction between these returns and the economic
variables. This analysis enables researchers to identify if the listed property indices returns
respond positively or negatively to one unit of information shocks from economic variables
(Downs et al. 2003; Brooks & Tsolacos 1999).
This test also reveals how quickly the returns absorb the information from changes in economic
indicator. The impacts of economic variables are the most significant in the first quarter
enduring for three quarters.
See Table 29 and Figure 11 for the results.
83
Table 29: Impulse response (Table format)
Period LB ALSI CPI J256T
1 -0.07 0.02 -0.01 0.07
2 0.01 0.01 -0.01 -0.01
3 0.02 0.02 0.01 -0.01
4 0.00 0.00 0.00 0.00
5 -0.01 0.00 0.00 0.00
6 0.00 0.00 0.00 0.00
7 0.00 0.00 0.00 0.00
8 0.00 0.00 0.00 0.00
9 0.00 0.00 0.00 0.00
10 0.00 0.00 0.00 0.00
The shock to interest rate had a largely negative response on listed properties returns for the
first two quarters; this was followed by three quarters of a slightly positive response. Then
beyond that it seems to work the shock appears to have worked its way out of the system. This
result is consistent with the findings of Ewing and Payne (2005)24, and similar to those of
Glascock, Lu & So (2002)25 and Payne (2003)26. The shock to stock market has a positive
response on properties returns; this result is in line with the findings of Payne (2003)27 and
Yunus (2012)28.
24 Ewing and Payne (2005) showed the shock was negative for first six months and positive subsequently. 25 Glascock, Lu and So (2002) found that the shock results in initially two months of negative effects and five months of positive effects. 26 Payne (2003) showed that the effect is negative and last six months. 27 Payne (2003) established that the impact is positive and last five months. 28 Yunus (2012) indicated that the impact is positive and last 15 months for the US market.
84
The shock to CPI had initially two quarters of negative response then in the third quarter it had
positive response. This result is similar to the findings of Glascock, Lu & So (2002)29, Payne
(2003)30 and Ewing and Payne (2005)31.
The shock to properties return (own shock) had the largest positive response for all variables
tested in the first quarter, followed by two quarters of negative response. The response is
similar to the findings by Glascock, Lu & So (2002)32 and Laopodis (2009)33.
Other than the CPI, the results from impulse response are consistent with regression analysis.
The longevity of the shock is similar to Brooks and Tsolacos (1999), where they found the shock
persists past 24 months (in the UK market). Glascock, Lu & So (2002)’s results showed the shock
persists after 15 months.
29Glascock, Lu and So (2002) found that the shock in the first month is zero and then positive for the next 14 months. 30 Payne (2003) found that the shock is negative for first six months and then reverts to zero. 31 Ewing and Payne (2005) found that the shock is negative for first six months and then reverts to zero. 32 Glascock, Lu and So (2002) indicated that the own shock is mostly positive and has the biggest impact relative to other variables. 33 Laopodis (2009) found that the own shock is positive and has the biggest impact relative to other variables tested.
85
Figure 11: Impulse response (Graph format)
The horizon axis is measured in quarters; the vertical axis measures the magnitude of the
response, scaled such that 1.0 equals 1 Standard Deviation. Confidence bands, used to
determine statistical significance, are shown as red dashed (——) lines and represent 2
Standard Errors.
-.10
-.05
.00
.05
.10
1 2 3 4 5 6 7 8 9 10
Response of J256T to LB
-.10
-.05
.00
.05
.10
1 2 3 4 5 6 7 8 9 10
Response of J256T to ALSIT
-.10
-.05
.00
.05
.10
1 2 3 4 5 6 7 8 9 10
Response of J256T to CPI
-.10
-.05
.00
.05
.10
1 2 3 4 5 6 7 8 9 10
Response of J256T to J256T
Response to Cholesky One S.D. Innovations ± 2 S.E.
86
4.3 Summary and discussion
Table 30 summarises the statistical tests conducted and the significant results from the tests.
Table 30: Summary of results
Unlisted Properties Listed Properties
Data – Sample period
analysis
18 annual data points (1995 to
2012) cover one and half
business cycle but not one full
interest rate cycle
46 quarterly data points (2002
– 2013) cover one and half
business cycle but not one full
interest rate cycle
Data – Histogram No outlier was detected and
none were removed from the
data set
No outlier was detected and
none were removed from the
data set
Data – Descriptive statistics No unusual data identified in
the descriptive statistics
No unusual data identified in
the descriptive statistics
Data – Jarque–Bera test The data is normally
distributed
The data is normally
distributed
Data – Correlation matrix IPD returns are positively
correlated to ALSI, CPI, and
GDP and negatively related to
the ten-year bond rate. No
concern with multicollinearity.
J256T returns are positively
correlated to ALSI and
negatively related to GDP, CPI
and ten-year bond rate. No
concern with
multicollinearity.
Regression – Results Unable to perform test due to
limited size of sample
ten-year bond rate was the
only variable to be found to
be statistically significant,
with negative coefficient.
Regression – Assumptions Unable to perform test due to
limited size of sample
All assumptions for multiple
regression met.
VAR – Histogram Unable to perform test due to
limited size of sample
No outlier was detected and
none were removed from the
data set
VAR – Dickey–Fuller test Unable to perform test due to
limited size of sample
GDP had to be removed from
the data set as it had unit
root.
VAR Unable to perform test due to
limited size of sample
No statistically significant
variable. All assumptions for
VAR are met, thus results are
87
valid.
VAR – Variance
decomposition
Unable to perform test due to
limited size of sample
Volatility of listed property
returns is most influenced by
interest rate and its own
volatility.
VAR – Impulse Response Unable to perform test due to
limited size of sample
The shock to interest rate has
large negative response to
listed property returns.
Unlisted properties
IPD returns are positively correlated to ALSI, CPI, and GDP and negatively related to the ten-
year bond rate. However, due to the limited data available for unlisted properties, detailed
statistical analysis (regression and VAR) was not performed on unlisted properties. Thus the null
hypothesis as stated in section 1.10 cannot be rejected.
Listed properties
Interest rate was the only statistically significant variable identified via regression model; all
other variables were not statistically significant.
Stock market (see section 4.2.3, Table 21) was significant at 8% level, just above the 5%
significance level. Thus, it can be concluded that the stock market is close to being considered
but is not statistically significant. Despite not being considered statistically significant, the
positive co-efficient (regression) and effect (impulse response) is compatible with most of the
previous research (see Allen, Madura & Springer 2000; Okunev, Wilson & Zurbruegg 2000;
Payne 2003; He, Webb & Myer 2003; Huang & Lee 2009; Chen et al. 2012; Yunus 2012).
Economic growth (see section 4.2.3, Table 21) was found not to be a significant variable. This is
similar with previous literature by Brooks and Tsolacos (1999), Standish et al. (2005) and Chang,
Chen & Leung (2011).
88
Inflation (see section 4.2.3, Table 21) was not a significant macroeconomic variable to property
returns. This is comparable with the majority of prior research cited in this study (Chen, Hsieh &
Jordan 1997; Chatrath & Liang 1998; Chen et al. 1998; Glascock, Lu & So 2002).
Interest rate (see section 4.2.3, Table 21, Table 22 and Table 23) was statistically significant
and was negatively related to property returns. This result is consistent with the theoretical
arguments and prior research (see Allen, Madura & Springer 2000; He, Webb & Myer 2003;
Clark & Daniel 2006; Huang & Lee 2009; Mangani 2011; Boshoff & Cloete 2012; Chen et al.
2012; Nittayagasetwat & Buranasiri 2012).
Variance decomposition (see section 4.2.4.1, Table 28) indicated that interest rates are the
greatest source of volatility. The interest rate variable explained almost 49% of the volatility of
property returns. The second largest contributor is own shock accounting for 48% of the
volatility of property returns, similar to results from Brooks and Tsolacos (1999), Downs et al.
(2003) and Laopodis (2009).
The impulse response result (see section 4.2.4.2, Table 29) is consistent with that of Ewing and
Payne (2005) that the shock to interest rate leads to negative returns. The impulse response
analysis on the stock market is in line with the findings of Payne (2003), Laopodis (2009) and
Yunus (2012) that property negatively responds to a shock to the stock market. The shock to
CPI had initially two quarters of negative response then in the third quarter it had a positive
response. This result is similar to the findings of Glascock, Lu and So (2002), Payne (2003), and
Ewing and Payne (2005). The shock to properties return (own shock) has the largest positive
response for all variables tested in the first quarter, then followed by two quarters of negative
response. The response is similar to the findings by Glascock, Lu and So (2002), and Laopodis
(2009). The results from impulse response are consistent with regression analysis.
Despite the lack of data for unlisted property, the strong correlation of 69% (see section 3.2.3)
between unlisted property (IPD) and listed property (J256T), can infer similar results for the
period studied.
89
Thus, based on statistical methods applied by this research, interest rate is a noteworthy
macroeconomic variable (and negatively related) to listed and unlisted properties for the period
studied. All other macroeconomic variables studied such as stock market, economic growth and
inflation were not found to be statistically significant; thus, no relationship between the
macroeconomic variables and commercial property returns can be concluded. Thus, the null
hypothesis as stated in section 1.10 can be rejected for interest rate. There is a statistically
significant relationship between interest rate and commercial properties returns (listed and
unlisted) returns.
90
5. Conclusion and Suggested Future Research
This research attempted to answer two questions (as stated in section 1.6):
a) Is there a significant relationship between macroeconomic variables and commercial
property returns (listed and unlisted) in South Africa?
b) If such a relationship exists, what are the relationships between the chosen
macroeconomic variables and commercial property returns?
In order to answer the two questions, this study reviewed and summarised previous local and
international literature on this subject matter. Four macroeconomic variables were considered
critical, namely: stock market performance (the majority of the studies found a positive
relationship between property returns and stock market performance); economic growth
(despite mixed results, most found a positive relationship consistent with economic theory);
interest rate (the majority of the studies found a negative relationship consistent with
theoretical argument); and inflation (most past studies found no significant relationship
between inflation and property returns).
Further, statistical methodologies – namely, cross-sectional regression and vector
autoregression (VAR) – were adopted from prior studies for this research. Relevant proxies for
the four variables were identified from prior literature and consideration was given to the
South African context. Data for the relevant proxies were collected and statistical analysis was
performed.
The statistical test result on unlisted property was found to be inconclusive due to insufficient
data.
The statistical tests on listed property demonstrated that interest rate is significantly negatively
related to listed property returns (using cross-sectional regression). The negative relationship
was further confirmed using impulse response analysis (VAR); the interest rate was also found
to be the largest contributor (almost 49%) to volatility of listed property returns (using variance
91
decomposition of VAR). The result of a negative relationship between the interest rate and
listed property returns is consistent with the majority of prior studies cited (see Allen, Madura
& Springer 2000; He, Webb & Myer 2003; Clark & Daniel 2006; Huang & Lee 2009; Mangani
2011; Boshoff & Cloete 2012; Chen et al. 2012; Nittayagasetwat & Buranasiri 2012).
The stock market was found to be close to being significant and positively related to property,
compatible with prior studies. Economic growth and inflation were not found to be significant
variables similar to prior international studies.
Thus, the null hypothesis (in section 1.10) for unlisted commercial property cannot be rejected
due to lack of data, and this study concludes that there is no statistically significant relationship
between macroeconomic variables and unlisted commercial properties for the period studied.
For listed commercial property, the null hypothesis (in section 1.10) can be rejected, and this
study concludes that there is a negative statistical significant relationship between interest rate
and listed property returns for the period studied.
The major obstacle to this research was the lack of available data (the lack of data resulted in
inconclusive result for unlisted properties). Thus, one area for future research is to repeat this
research using a larger dataset, especially covering more than one full interest-rate cycle. Other
potential research areas would be to perform this analysis by splitting the returns into income
returns and price returns individually, as conducted by Downs et al. (2003).
92
6. References and Bibliography
Adrangi, B, Chatrath, A & Raffiee, K 2004, ‘REIT Investments and hedging against Inflation’, Journal of Real Estate Portfolio Management, vol. 10, no. 2, pp. 1–10.
Allen, M, Madura, J & Springer, T 2000, ‘REIT Characteristics and the Sensitivity of REIT Returns’. Journal of Real Estate Financial Economics, vol. 21, no. 2, pp. 141–152.
Anderson, A 2013a, ‘Demand for residential properties attracting investors’, Business Day, 13 November, viewed 6 December 2013, http://www.bdlive.co.za/business/property/2013/11/13/demand-for-residential-properties-attracting-investors
Anderson, A 2013b, ‘Arrowhead acquires 31.7% of Vividend Income Fund units’, Business Day, 6 December, viewed 6 December 2013, http://www.bdlive.co.za/business/property/2013/12/06/arrowhead-acquires-31.7-of-vividend-income-fund-units
Appraisal Institute 2008, The Appraisal of Real Estate, Thirteenth Edition, Chicago.
Blackrock Real Estate Equity Group 2005, Real Estate Performance in a Rising Interest Rate Environment: An Empirical Analysis, viewed 1 December 2013, http://www.rmortgage.com/Images/docs/InterestRateAnalysis2-04.pdf
Boshoff, D & Cloete, C 2012, ‘Can Listed Property shares be a surrogate for direct property investment behavior?’ South Africa Journal of Economics and Management, vol. 15, no. 1, pp. 72–91.
Brooks, C & Tsolacos, S 1999, ‘The impact of economic and financial variables on UK property performance’, Journal of Property Research, vol. 16 no. 2, pp. 139–152.
Boudry, W, Coulson, E, Kallberg, J & Liu, C 2012 ‘On the Hybrid Nature of REITs’, Journal of Real Estate Financial Economics, vol. 44, no. 1, pp. 230–249, doi: 10.1007/s11146-011-9339-7
Bureau of Economic Research 2013, email, received 7 October from [email protected], University of Stellenbosch, http://www.ber.ac.za
Business Day 2013, ‘Company and Markets’, viewed 26 November, http://www.bdlive.co.za/markets/
CBRE Clarion Securities 2013, ‘REIT market roundtable Q&A: How are REITs positioned in a rising interest rate environment’, viewed 1 December 2013, http://www.cbreclarion.com/research/Documents/CBRE%20Clarion%20Interest%20Rates%20and%20REITS%20-%20Summer%202013.pdf
93
Chang, K, Chen, N & Leung, CKY 2011, ‘Monetary Policy, Term Structure and Asset Return: Comparing REIT, Housing and Stock’, Journal of Real Estate Financial Economics, vol. 43, pp. 221–257, doi: 10.1007/s11146-010-9241-8
Chatrath, A & Liang, Y 1998, ‘REITs and Inflation: A Long-Run Perspective’, Journal of Real Estate Research, vol. 16, no. 3, pp. 311–326.
Chen, S, Hsieh, C & Jordan B 1997, ‘Real Estate and the Arbitrage Pricing Theory: Macrovariables vs Derived Variables’, Real Estate Economics, vol. 25, no. 3, pp. 505–523.
Chen, S, Hsieh, C, Vines, TW & Chiou, S 1998, ‘Macroeconomic Variables, Firm-Specific Variables and Returns to REITs’, Journal of Real Estate Research, vol. 16, no 3, pp. 269–277.
Chen, M, Peng, C, Shyu, S & Zeng, J 2012, ‘Market States and the Effect on Equity REIT Returns due to Changes in Monetary Policy Stance’. Journal of Real Estate Financial Economics, vol. 45, pp. 364–382, doi: 10.1007/s11146-010-9264-1
Chen, N, Roll, R & Ross, SA 1986, ‘Economic forces and the stock market’, Journal of Business, vol. 59, pp. 383–403.
Clark, AE & Daniel T 2006, ‘Forecasting South African house prices’, Investment Analysts Journal, no. 64, pp. 27–34.
Clayton, J & MacKinnon, G 2001, ‘The Time-Varying Nature of the link between REIT, Real Estate and Financial Returns’, Journal of Real Estate Portfolio Management, vol. 7, no. 1, pp. 43–54.
Cohen & Steers Capital Management 2013, ‘What History Tells Us About REITs, Inflation and Rising Rates’, viewed 1 December 2013, http://www.cohenandsteers.com/assets/content/resources/insight/What_History_Tells_Us_About_REITs_Inflation_and_Rising_Rates.pdf
Das, S, Gupta, R, Kanda, P, Tipoy, C & Zerihun, M 2012, ‘Real Interest Rate Persistence in South Africa: Evidence and Implications’, Working Paper 2012-04, viewed 1 January 2013, http://web.up.ac.za/sitefiles/file/40/677/WP_2012_04.pdf
DiPasquale, D & Wheaton, WC 1992, ‘The Markets for Real Estate Assets and Space: A Conceptual Framework’, Journal of American Real Estate and Urban Economics Association, vol. 20, no. 1, pp. 181–197.
Downs, DH, Fung, H, Patterson, GA & Yua, J 2003, ‘The Linkage of REIT Income- and Price-Returns with Fundamental Economic Variables’, Journal of Alternative Investments, vol. 6, no. 1, pp. 39–50.
Du Toit, H & Cloete, C 2004, ‘Appraisal of the Fischer-DiPasquale-Wheaton (FDW) Real Estate Model and Development of an Integrated Property and Asset Market Model’, South African Journal of Economic and Management Sciences, vol. 7, no. 2, pp. 341–367.
94
Ewing, BT & Payne, JE 2005, ‘The response of real estate investment trust returns to macroeconomic shocks’, Journal of Business Research, vol. 58, pp. 293–300.
Franken, M, Bloom, J & Erasmus, P 2011, ‘Factors that affect South African real estate price growth’, Management Dynamics, vol. 20, no. 2, pp. 18–32.
Ferguson, N 2008, The Ascent of Money: A Financial History of the World, The Penguin Press, New York.
Firer, C & McLeod, H 1999, ‘Equities, bonds, cash and inflation: historical performance in South Africa 1925 to 1998‘, Investment Analysts Journal, no. 50, pp. 7–28.
Graff, R 2001, ‘Economic Analysis suggests that REIT Investment characteristics are not as advertised’, Journal of Real Estate Portfolio Management, vol. 7, no. 2, pp. 99–124.
Glascock, JL, Lu, C & So, RW 2002, ‘REIT Return and Inflation: Perverse or Reverse Causality Effects?’ Journal of Real Estate Finance and Economics, vol. 24, no. 3, pp. 301–17.
Gujarati, DN 2003, Basic Econometrics, 4th Edition, McGraw Hill, New York.
Hassan, S & Biljon, A 2009, ‘The Equity Premium and Risk-Free Rate Puzzles in a Turbulent Economy: Evidence from 105 Years of Data from South Africa’, Working Paper 156, Economic Research Southern Africa, viewed 1 January 2013, http://www.econrsa.org/
INET BFA 2013, Data obtained from database, viewed 1 June – 31 Dec 2013, http://www.inetbfa.com/
He, LT, Webb, JR & Myer, FCN 2003, ‘Interest Rate Sensitivities of REIT Returns’, International Real Estate Review, vol. 6, no. 1, pp. 1–21.
Hedley, N 2013, ‘Unlisted property sector “holds some of SA’s most prized assets”’, Business Day, 24 May, viewed 1 December 2013, http://www.bdlive.co.za/business/property/2013/05/24/unlisted-property-sector-holds-some-of-sas-most-prized-assets
Hoesli, M & MacGregor, B 2000, Property Investment: Principles and Practice of Portfolio Management, Pearson Education, Edinburgh.
Hoesli, M, Lizieri, C & MacGregor, B 2008, ‘The Inflation Hedging Characteristics of US and UK Investments: A Multi-Factor Error Correction Approach’, Journal of Real Estate Financial Economics, vol. 36, pp. 183–206, doi: 10.1007/s11146-007-9062-6
Huang, C & Lee, Y 2009, ‘The relationship between Oil Price Growth and REIT Returns’, International Research Journal of Finance and Economics, no. 33, pp. 120–133.
Ilmanen, A 2012, Expected Returns on Major Asset Classes, Wiley and Sons, Inc., Chichester, ISBN 978-1-934667-48-4
95
International Valuation Standards Council 2011, International Valuation Standards, viewed 1 December 2013, http://www.ivsc.org/
JP Morgan Asset Management 2013, ‘What will higher interest rates mean for real estate?’ viewed 1 December 2013, http://www.jpmorganinstitutional.com/blobcontent/520/859/1323363596727_II_Real%20Estate%20What%20Higher%20Interest%20Rates%20Mean_r9.pdf
Laopodis, N 2009, ‘REITs, the stock market and economic activity’, Journal of Property Investment and Finance, vol. 27, no. 6, pp. 563–578.
Lieser, K & Groh, AP 2013, ‘The Determinants of International Commercial properties Investment’, Journal of Real Estate Financial Economics, vol. 48, no. 4, pp. 611–659, doi: 10.1007/s11146-012-9401-0
Lieser, K & Groh, AP 2011, ‘The Attractiveness of 66 Countries for Institutional Real Estate Investments’, Journal of Real Estate Portfolio Management, vol. 17, no. 3, pp. 1–17.
Ling, D & Naranjo, A 1997, ‘The Economic Risk Variables and Commercial properties Returns’, Journal of Real Estate Finance and Economics, vol. 15, no. 3, pp. 283–307.
Liow, K 2010, ‘Integration between Securitised Real Estate and Stock Markets: A Global Perspective‘, Journal of Real Estate Portfolio Management, vol. 16, no. 3, pp. 249–265.
Lynn, D 2007, ‘The Tectonic Forces of Global Real Estate: Implications for Global Investments and Portfolio Managers’, Journal of Real Estate Portfolio Management, vol. 13, no. 1, pp. 87–92.
Mangani, R 2011, ‘Monetary Policy, structural breaks and JSE Returns’, Investment Analysts Journal, no. 73, pp. 27–35.
Mueller, GR & Pauley KR 1995, ‘The effect of interest rate movements on Real Estate Investment Trust’, The Journal of Real Estate Research, vol. 10, no. 3, pp. 319–325.
Nittayagasetwat, A & Buranasiri, J 2012, ‘Real Estate Investment Performance: the Test of the Impact of Additional Interest Rate Information from CIR Model’, International Journal of Business and Social Science, vol. 3, no. 12, pp. 134–143.
Okunev, J, Wilson, P & Zurbruegg, R 2000, ‘The Causal Relationships between Real Estates and Stock Markets’, Journal of Real Estate Finance and Economics, vol. 21, no. 3, pp. 251– 261.
Payne, JE 2003, ‘Shocks to macroeconomic state variables and the risk premium of REITs’, Applied Economics Letters, no. 10, pp. 671–677, doi: 10.1080/1350485032000133345
Pindyck, R & Rubinfeld, D 1998, Econometric Models and Economic Forecasts, Fourth Edition, McGraw-Hill, Singapore.
96
Property Sector Charter Council 2012, ‘New research values South Africa's property sector at R4.9 trillion’, Property Charter: Latest News, viewed 1 June 2014, http://www.propertycharter.co.za/news_article.htm?contentID=7
PwC (PricewaterhouseCoopers) 2012, An African perspective: Valuation methodology survey 2012, viewed 1 November 2013, http://www.pwc.co.za/en_ZA/za/assets/pdf/valuation-methodology-survey-2012.pdf
PwC (PricewaterhouseCoopers) 2010, Signs of the times: Valuation Methodology Survey 2010, viewed 1 November 2013, http://www.pwc.co.za/en/assets/pdf/pwc-valuation-methodology-survey-2010.pdf
SA Corporate Real Estate Fund (no date), ‘Investor FAQs’, viewed 1 October 2013, http://www.sacorporatefund.co.za/investor-relations/investor-faqs/
SA REIT Association 2013a, ‘Chart Book – August 2013’, viewed 1 October 2013, http://www.sareit.com/stats.php
SA REIT Association 2013b, ‘SA REIT Association Launch – 23 May 2013’, viewed 1 December 2013, http://www.sareit.com/docs/SAREIT_Presentation – 23_May_2013.pdf
SA REIT Association 2013c, ‘Chart Book – October 2013’, viewed 1 December 2013, http://www.sareit.com/stats.php
Schnehage, M 2012, ‘Foreign investors on steroids for SA listed property’, Moneyweb, 1 August, viewed 1 June 2014, http://www.moneyweb.co.za/moneyweb-property/foreign-investors-on-steroids-for-sa-listed-proper
Simo-Kengne, B, Bittencourt, M & Gupta, R 2012, ‘House Prices and Economic Growth in South Africa: Evidence from Provincial-Level Data’, Journal of Real Estate Literature, vol. 20, no. 1, pp. 97–117.
Simo-Kengne, B, Bittencourt, M & Gupta, R 2013, ‘Macro shocks and House prices in South Africa’, Working Paper, Department of Economics, University of Pretoria.
Simpson, MW, Ramchander, S & Webb, JR 2007, ‘The Asymmetric Response of Equity REIT Return to Inflation’, Journal of Real Estate Financial Economics, vol. 34, no. 4, pp. 513–529, doi: 10.1007/s11146-007-9023-0
Sims, C 1980, ‘Macroeconomics and Reality’, Econometrica, vol. 48, no. 1, pp. 1–48.
Smith, C 2013, ‘REITs Law Sparks Foreign Investor Interest in SA Property’, South African Institute of Tax Professionals (SAIT), 17 December, viewed 1 June 2014, http://www.thesait.org.za/news/150217/REITs-Law-Sparks-Foreign-Investor-Interest-in-SA-Property.htm
South African Government 2013, Section 25BB of the Taxation Legislation Amendment Bill, viewed 1 January 2014, http://www.treasury.gov.za
97
Standish, B, Lowther, B, Morgan-Grenville, R & Quick, C 2005, ‘The determinants of residential house prices in South Africa’, Investment Analysts Journal, no. 61, pp. 41–48.
Stock, JH & Watson, MW 2001, ‘Vector Autoregressions’, Journal of Economic Perspectives, vol. 15, no. 4, pp. 101–115, doi: 10.1257/jep.15.4.101
Swanson, Z, Theis, J & Casey KM 2002, ‘REIT Risk Premium Sensitivity and Interest Rates’, Journal of Real Estate Finance and Economics, vol. 24, no. 3, pp. 319–330.
Towers Watson 2012, ‘Expected property returns when government bond yields are rising’, viewed 1 December 2013, http://www.towerswatson.com/en-GB/Insights/IC-Types/Ad-hoc-Point-of-View/2012/Expected-property-returns-when-government-bond-yields-are-rising
UCT (University of Cape Town) 2011, CON5041Z: Introduction to Applied Statistics, Department of Statistics, Cape Town.
Wooldridge, JM 2012, Introductory Econometrics – A Modern Approach, Fifth Edition, South-Western Cengage Learning, Mason, Ohio.
Yobaccio, E, Rubens, J & Ketcham, D 1995, ‘The inflation-hedging properties of risk assets: the case of REITs’, Journal of Real Estate Research, vol. 10, no. 3, pp. 279–296.
Yunus, N 2012, ‘Modeling Relationships among securitized property market, stock markets and macroeconomic variables’, Journal of Real Estate Research, vol. 34, no. 2, pp. 127–156.
98
Appendices Cover Page
Appendix 1: Investment analysts’ views on listed property
Appendix 2: Data used
99
Appendix 1: Investment analysts’ views on listed
property
Table 31 summarises a brief survey of investment analysts indicates that listed property is
highly correlated to long-term government bonds or used as a benchmark to determine the
attractiveness of listed property as an asset class (see section 2.3.1, under investments rate vs
borrowing rate).
Table 31: Summary of survey of analysts
Asset
Manager
View Date and Source
Sanlam
(Rafiq Taylor)
‘The volatility can largely be attributed to
movements in bond yields and the rand. We expect
listed property to remain volatile in the short term
due to its strong correlation with the bond market.’
‘…the current interest rate cycle, as bond yields and
property yields are highly correlated.’
18 July 2013
https://www.sanlam.co.za/wps/
wcm/connect/sanlam_en/sanla
m/media+centre/media+release
s/volatility+supports+a+wait-
and-
see+approach+to+listed+proper
ty
Catalyst
‘…With the listed property sector’s yields having
rerated relative to the government bond index by
0.63%, “the listed property historic rolled yield is
now trading at a premium spread to the long-term
government bond index yield of 1.25%, compared
with the five-year average of 0.18%...”’
10 July 2013
http://www.bdlive.co.za/busine
ss/property/2013/07/10/correla
tion-between-bonds-and-listed-
property-weakening
Grindrod
(Ian
Anderson)
‘…the South African listed property sector was
trading on a forward yield of 6.9% — about 90 basis
points above the yield on a ten-year government
bond. This leaves the sector vulnerable to further
weakness in global and local bond markets and is
likely to result in significant price volatility in the
short term…’
10 July 2013
http://www.bdlive.co.za/busine
ss/property/2013/07/10/correla
tion-between-bonds-and-listed-
property-weakening
100
Plexus Asset
Management
(Paul
Stewart)
‘...correlations between the bond (ALBI) and real
estate (SAPI) markets proved to be much greater
than those witnessed between equities and real
estate...’
‘Between 49% to 94%’
11 August 2011
http://www.moneyweb.co.za/m
oneyweb-property/is-listed-
property-correlated-more-with-
bonds-or-e
Coronation
(Anton De
Goede)
‘…over the past ten years, the local listed property
sector has really benefitted from ... the subsequent
rerating in the local bond market in general.’
‘…the derating in the property market following the
spike in bond yields is related to the sector’s
perceived yield prospects.’
10 December 2013
http://www.moneyweb.co.za/m
oneyweb-property/listed-
property-in-2014
Old Mutual
Properties
(Peter
Levett)
‘While property and bond yields generally trend
together over the long term…’
‘long-term correlation of 89%’
http://www.oldmutual.co.za/do
cuments/Insights/PropMktStrat.
Erwin Rode Erwin Rode, property valuer and economist at Rode
and Associates, says, ‘the correlation between long
bond yields and listed property yields is strange,
since property’s income stream grows with about
6% on average while that of bonds is fixed.
It is therefore odd that the market acts on the
perceived similarities between the asset classes.
However, one cannot argue with the market and
has to accept its behaviour.’
10 December 2013
http://www.moneyweb.co.za/m
oneyweb-property/listed-
property-in-2014
ABSA
(Mariette
Warner)
‘From 2003 to 2005, the SAPY increased by 113%
(28.7% p.a.) because of falling long bond yields –
from 10% to 7.5%.’
http://www.bondstreet.co.za/p
dfs/desk/Listed%20Property%20
Historical%20Perspective.pdf
Prudential ‘…property yields are still out of line with bond
yields, having sold off less than bonds so far this
20 October 2013
101
year…’ http://www.iol.co.za/business/p
ersonal-finance/financial-
planning/investments/bad-
quarter-for-listed-property-
1.1594309#.UrskKfQW2So
102
Appendix 2: Data used
The raw data used in this research are in the tables below. (A softcopy of the data is included in
an USB flash drive as part of the final submission.)
Data set for section 4.1
Annual
Stock Market Economic Inflation
Interest Rate
IPDT ALSI GDP CPI LB
1994
0.0320 0.0896 0.1483
1995 0.1530 0.0819 0.0310 0.0869 0.1611
1996 0.1400 0.0953 0.0430 0.0735 0.1548
1997 0.1740 (0.0691) 0.0260 0.0860 0.1470
1998 0.0500 (0.0591) 0.0050 0.0687 0.1512
1999 0.1360 0.7082 0.0240 0.0520 0.1490
2000 0.1100 0.0035 0.0420 0.0533 0.1379
2001 0.1040 0.3261 0.0270 0.0571 0.1141
2002 0.0950 (0.0831) 0.0370 0.0915 0.1150
2003 0.1520 0.1608 0.0290 0.0588 0.0962
2004 0.2340 0.2544 0.0460 0.0138 0.0953
2005 0.3000 0.4725 0.0530 0.0340 0.0807
2006 0.2740 0.4123 0.0560 0.0464 0.0794
2007 0.2770 0.1919 0.0550 0.0710 0.0799
2008 0.1280 (0.2323) 0.0360 0.1153 0.0910
2009 0.0910 0.3213 (0.0150) 0.0712 0.0870
2010 0.1340 0.1898 0.0310 0.0427 0.0862
2011 0.1040 0.0257 0.0350 0.0500 0.0852
2012 0.1520 0.2668 0.0250 0.0565 0.0790
Data set for section 4.2
Listed Properties
Stock Market Economic Inflation
Interest Rate
J256T ALSI GDP CPI LB
30-Sep-02 0.03266 -0.10111 -0.0040 0.0268 -0.0041
31-Dec-02 0.12736 -0.01350 -0.0015 0.0240 -0.0034
31-Mar-03 0.12260 -0.16275 0.0028 -0.0210 -0.0082
30-Jun-03 0.06479 0.09684 -0.0045 -0.0290 -0.0039
103
30-Sep-03 -0.02280 0.07932 0.0005 -0.0310 -0.0006
31-Dec-03 0.19727 0.17118 0.0002 -0.0400 -0.0040
31-Mar-04 -0.00635 0.03676 0.0098 -0.0030 0.0024
30-Jun-04 0.03757 -0.04723 -0.0013 0.0030 0.0073
30-Sep-04 0.11509 0.17433 0.0025 0.0060 -0.0046
31-Dec-04 0.22415 0.08139 -0.0060 0.0190 -0.0086
31-Mar-05 0.06305 0.05949 -0.0005 -0.0030 -0.0075
30-Jun-05 0.12530 0.07243 0.0083 0.0030 0.0022
30-Sep-05 0.16421 0.20286 -0.0045 0.0070 -0.0029
31-Dec-05 0.11638 0.07740 -0.0073 -0.0020 -0.0017
31-Mar-06 0.22128 0.13252 0.0088 0.0010 -0.0052
30-Jun-06 -0.17217 0.04866 0.0013 0.0020 0.0046
30-Sep-06 0.09932 0.06334 -0.0023 0.0120 0.0081
31-Dec-06 0.20344 0.11834 0.0015 0.0030 -0.0058
31-Mar-07 0.15951 0.10381 0.0003 0.0040 -0.0044
30-Jun-07 0.01157 0.04288 -0.0085 0.0110 0.0019
30-Sep-07 0.11171 0.06713 0.0048 0.0000 0.0061
31-Dec-07 -0.01242 -0.02969 0.0025 0.0140 -0.0019
31-Mar-08 -0.09836 0.02918 -0.0075 0.0150 0.0054
30-Jun-08 -0.19679 0.03361 0.0035 0.0170 0.0093
30-Sep-08 0.24529 -0.20555 -0.0065 0.0180 -0.0022
31-Dec-08 0.08302 -0.09166 -0.0088 -0.0230 -0.0090
31-Mar-09 -0.01096 -0.04194 -0.0115 -0.0270 -0.0037
30-Jun-09 -0.02168 0.08650 0.0090 -0.0070 0.0054
30-Sep-09 0.12538 0.13906 0.0110 -0.0130 0.0015
31-Dec-09 0.03412 0.11441 0.0045 -0.0040 0.0018
31-Mar-10 0.10089 0.04479 0.0023 -0.0030 0.0004
30-Jun-10 0.00562 -0.08173 -0.0033 -0.0120 -0.0019
30-Sep-10 0.14455 0.13291 0.0013 -0.0100 -0.0061
31-Dec-10 0.03344 0.09470 0.0020 0.0000 -0.0011
31-Mar-11 -0.02311 0.01116 0.0010 0.0030 0.0054
30-Jun-11 0.05072 -0.00608 -0.0073 0.0080 -0.0012
30-Sep-11 0.01840 -0.05836 0.0000 0.0080 -0.0030
31-Dec-11 0.03058 0.08380 0.0035 0.0070 0.0016
31-Mar-12 0.09117 0.06005 -0.0020 0.0000 -0.0013
30-Jun-12 0.12001 0.00979 0.0023 -0.0040 -0.0009
30-Sep-12 0.11491 0.07257 -0.0055 -0.0060 -0.0077
31-Dec-12 0.02100 0.10338 0.0023 0.0050 0.0009
